ENFSI-BPM-DI-01 (vs.

01) BPM for Facial Image Comparison

Best Practice Manual

for Facial Image Comparison
ENFSI-BPM-DI-01
Version 01 - January 2018
ENFSI’s position on Best Practice Manuals
ENFSI wishes to promote the improvement of mutual trust by encouraging forensic
harmonisation through the development and use of Best Practice Manuals.
Furthermore, ENFSI encourages sharing Best Practice Manuals with the whole Forensic
Science Community which also includes non ENFSI Members.
Visit www.enfsi.eu/documents /bylaws for more information. It includes the ENFSI policy
document Policy on Creation of Best Practice Manuals within ENFSI (code: QCC-BPM-001).

Official language
The text may be translated into other languages as required. The English language version
remains the definitive version.

Copyright
The copyright of this text is held by ENFSI. The text may not be copied for resale.

Further information
For further information about this publication, contact the ENFSI Secretariat. Please check the
website of ENFSI (www.enfsi.eu) for update information.
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

Best Practice Manual for Facial Image Comparison

CONTENTS

ACKNOWLEDGEMENTS .......................................................................................................... 4
1. AIMS ............................................................................................................................ 4
2. SCOPE ........................................................................................................................ 4
3. DEFINITIONS AND TERMS ........................................................................................ 5
4. RESOURCES .............................................................................................................. 6
5. METHODS ................................................................................................................... 9
6. VALIDATION AND ESTIMATING UNCERTAINTY OF MEASUREMENT ................ 16
7. PROFICIENCY TESTING .......................................................................................... 17
8. HANDLING ITEMS .................................................................................................... 17
9. INITIAL ASSESSMENT ............................................................................................. 17
10. PRIORITISATION AND SEQUENCE OF EXAMINATIONS ...................................... 18
11. RECONSTRUCTION OF EVENTS ............................................................................ 19
12. EVALUATION AND INTERPRETATION ................................................................... 19
13. PRESENTATION OF EVIDENCE .............................................................................. 20
14. HEALTH AND SAFETY ............................................................................................. 21
15. REFERENCES .......................................................................................................... 21
16. AMENDMENTS AGAINST PREVIOUS VERSION ................................................... 23
APPENDIX A: EDUCATION AND TRAINING GUIDELINES FOR FACIAL EXAMINERS ..... 24

APPENDIX B: FACTORS AFFECTING FACIAL APPEARANCE ........................................... 36

APPENDIX C: OVERVIEW OF COMPARISON METHODS ................................................... 48

Acknowledgements
Reuben Moreton (Metropolitan Police Service, UK), Sergio Castro Martinez (Cuerpo Nacional
de Policίa, Spain), Niclas Appleby, Fredrik Eklof, Elisabet Leitet, Klas Brorsson Läthén (National
Forensic Centre, Sweden), Alice Towler (University of New South Wales, Australia and
University of York, UK) and Arnout Ruifrok (Netherlands Forensic Institute, Netherlands).

Appendix A was produced as part of the CEPOL Knowledge Exchange Programme 2015
between Metropolitan Police Service, UK and National Forensic Centre, Sweden.

3/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

1. AIMS
This Best Practice Manual (BPM) aims to provide a framework for procedures, quality principles,
training processes and approaches to forensic facial image comparison (FIC). This BPM can be
used by member laboratories of the European Network of Forensic Science Institutes (ENFSI)
and other forensic science laboratories to establish and maintain working practices in the field
of forensic facial image comparison that will deliver reliable results, maximize the quality of the
information obtained and produce robust evidence. The use of consistent methodology and the
production of more comparable results will facilitate interchange of data between laboratories.

The term BPM is used to reflect the scientifically accepted practices at the time of creating. The
term BPM does not imply that the practices laid out in this manual are the only good practices
used in the forensic field. In this series of ENFSI Practice Manuals the term BPM has been
maintained for reasons of continuity and recognition.

2. SCOPE
2.1 General
This BPM addresses the processing, examination and comparison of imagery depicting faces
and the evaluation of findings in a forensic context.

This BPM is aimed at facial image examiners and assumes prior knowledge in the discipline of
facial comparison. It is not a standard operating procedure and addresses the requirements of
the judicial systems in general terms only.

BPMs are overarching document that describe recommended working practices for a process.
BPMs sit above detailed standard operating procedures, which describe the laboratory specific
method of a process. The recommendations within the BPM are not mandatory and will be
updated to reflect new empirical research where appropriate.

Facial comparison is widely used in other areas of law enforcement, such as when gathering
intelligence on persons of interest, during the process of a criminal investigation, verifying a
person’s identity for access control or reviewing the output of an automated facial recognition
system. These types of comparison are defined as facial review by the Facial Identification
Subcommittee of OSAC, in document ASTM E2916 Terminology for Digital and Multimedia
Evidence Examination [1] as:

facial review, (1) the review of a facial recognition (FR) system candidate list to
identify a possible match; (2) the one-to-one comparison (verification) conducted in a
high-throughput environment (for example, border crossing).

Generally facial reviews are considered to be non-evidential facial comparisons, staff who
conduct these comparisons may be trained facial reviewers but not to the level of facial
examiner. Although facial reviews are currently outside the scope of this BPM the principles
discussed for forensic facial image comparison are still relevant to facial review.

For the purpose of this document the term imagery will be used to describe both still images and
moving video.

Activity level reporting is outside the scope of this document.

4/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

2.2 Limitations
This BPM does not address the following aspects of facial identification:
• The recognition of familiar faces
• The use of automated facial recognition algorithms
• Age-progression of faces
• Facial reconstruction
• Facial composites
• Determining subject age and ethnicity from images
• Forensic artistry

This BPM also does not address comparison of other parts of the human body, object and
clothing comparison or gait analysis from imagery. Future versions of this BPM may address
other areas of comparison and also reviewing results from automated facial recognition systems.

The restoration of corrupt or incomplete digital data, physical repair of storage media, repair
of analogue media and complex imagery enhancement (e.g. frame integration techniques)
are outside the scope of this BPM. Where relevant and appropriate these tasks should be
undertaken by examiners with competency and training relevant to that field.

This BPM will address the authentication of imagery insofar as to determine if the imagery has
been provided in its original format. Authentication methods to determine if imagery has been
manipulated are outside the scope of this BPM.

3 DEFINITIONS AND TERMS

For the purposes of this BPM, the relevant terms and definitions given in ENFSI documents, the
International Laboratory Accreditation Cooperation (ILAC) G19 “Modules in a Forensic Science
Process” [ILAC G19:08/2014], as in standards like International Standardisation Organisation
(ISO) 9000 [ISO 9000:2005], ISO 17000 [ISO/IEC 17000:2004], 17020 [ISO/IEC 17020:2012]
and 17025 [ISO/IEC 17025:2005] apply.

For definitions of terms specific to facial image comparison used in this BPM refer to the
document: ASTM E2916 Terminology for Digital and Multimedia Evidence Examination [1].

For definitions of terms specific to image processing and enhancement used in this BPM please
refer to the ENFSI BPM for Forensic Image and Video Enhancement (ENFSI-BPM-DI-02) [2]
(Note that at the time of writing this BPM (ENFSI-BPM-DI-01), the BPM for Forensic Image
and Video Enhancement (ENFSI-BPM-DI-02) is in draft and a date of publication has not been
finalised).

For definitions and terms related to the evaluation of evidence and reporting conclusions refer to
the ENFSI Guideline for Reporting Evaluative Opinions [3].

Other specific terms used in this BPM are:

Questioned image: Imagery in which the identity of the depicted subject is contested or
uncertain. Questioned imagery is typically captured in uncontrolled conditions.

Reference image: Imagery in which the identity of the depicted subject is known and has been
verified. Reference imagery is often captured under controlled conditions.

5/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

4. RESOURCES

4.1 Personnel

4.1.1 Facial Image Examiners

This BPM is specific to facial examiners conducting FIC examinations. FIC examinations are
defined as follows by FISWG [4]:

‘The task of facial examination includes, but is not limited to, a rigorous one-to-
one analysis, comparison, and evaluation of controlled and uncontrolled images
for the purpose of effecting a conclusion. Examiners in this situation have to draw
on a larger foundation of knowledge, skill, and ability to accurately reach their
conclusions. Additionally, the articulation of the scientific and legal basis for the
expression of conclusions for many forensic, intelligence, or law enforcement
purposes requires an even more advanced level of training to include an expanded
set of knowledge, skills, and abilities above the level of basic concepts.’

Throughout this document the terms ‘facial image comparison’ and ‘FIC’ will be used to refer to
all aspects of the facial comparison examination.

4.1.2 Testing Innate Ability

The laboratory should implement a facial comparison testing programme as part of its
recruitment scheme, to gauge examiner ability before training. There is a wide range in innate
ability when comparing unfamiliar faces [5], [6]. Such a testing scheme can be used to highlight
individuals with high innate performance for recruitment as facial examiners and enrolment in
subsequent training and mentoring.

Throughout training regular testing and evaluation of trainee performance should be conducted
using ground truth facial comparison tasks. These tasks should cover the range of imagery that
will typically be encountered in casework. Empirical research has demonstrated that providing
trainees with feedback on facial comparison tests improves trainee ability [7].

4.1.3 Training
In order to undertake FIC, examiners must be trained and competent in appropriate procedures
and have specialist knowledge in relevant topics. The following is a list of recommended
competencies for facial examiners. Which competencies are required will depend upon the
operational requirements of the agency or organisation for whom the examiner works and the
types of casework undertaken. Not all of the points will be relevant to every agency or
organisation and not all topics are required to become a competent facial examiner in a
specific situation.

A detailed list of topics and references for the education and training of facial examiners can be
found in Appendix A.

New examiners should have a training plan devised prior to commencing training and be
assigned a mentor to guide them through their training and provide feedback on tasks. The
organisation’s Quality Management System (QMS) should describe how examiners are to be
trained and tested to achieve competency and how competency should be maintained.

6/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

4.2 Equipment
IT hardware and software is utilised for the viewing, processing and possible enhancement of
imagery for FIC. Such hardware and software should be fit for purpose and where possible
validated for its intended use. The laboratory should keep an up-to-date list of all hardware and
software used including identifying information such as the version or build number of software
and the serial number of hardware.

There is a wide range of software and hardware that can be used for facial image comparison,
these can be broadly defined as:

Hardware
• Computer hardware
• Storage and archiving system
• Graphical output devices (e.g. displays or printers)
• Graphical input devices for analogue media if appropriate (e.g. scanners, digital
cameras, video capture cards)

Graphical output devices, such as display monitors and printers, should be calibrated to
determine the accuracy of the outputted image. Likewise graphical input devices should also be
calibrated to ensure that the digital image generated captures information at the required level
of detail.

Software
• General purpose image processing tools
• General purpose video editing tools
• Image and video data analysis tools
• General and manufacturer specific (proprietary) viewing tools

The software used to view and process digital imagery should be validated insofar as possible
for its intended purpose. This validation should, at the bare minimum, determine if the process
results in any visual loss of quality to the displayed or outputted imagery and if this loss has an
impact upon the subsequent examination and comparison.

If it is not possible or appropriate to validate a piece of software, such as proprietary CCTV

replay software, examiners should use the ‘dual-tool’ approach and utilise more than one piece
of software to try to achieve consistent results [8].

4.3 Reference materials

The use of reference materials from a known source can be used to validate processes, test
examiner competency and calibrate equipment.

4.3.1 Validating technical processes

For image capture and enhancement techniques reference materials may be used for validating
the process, to ensure that the process is fit for purpose and produces the intended result. For
example, by using a reference signal or data input a process can be checked for any types of
degradation occurring in the signal pathway by observing the process output (See ENFSI BPM
for Forensic Image and Video Enhancement (ENFSI-BPM-DI-02) [2]).

4.3.2 Testing examiner competency

The laboratory should retain a collection of facial imagery that can be used for internal

7/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

proficiency testing of examiners and validation of comparison procedures. The ground truth
of this material should be known (i.e. the true responses for match and non-match pairs
are known) and encompass a range of image qualities that accurately represents casework
material. For details on conducting proficiency tests see ENFSI Guidance on the conduct of
Proficiency Tests and Collaborative Exercises [9]

Proficiency tests should be conducted using laboratory procedures to meaningfully test

examiner competency in those procedures. As comparison methods are subjective they are
reliant on the observations and interpretation of the examiner. FIC proficiency test should be
designed to allow effective evaluation of human performance [10].

4.4 Accommodation and environmental conditions

FIC should be conducted in a controlled environment. As with all forensic disciplines safe
and secure storage of submitted items is required to ensure the integrity and continuity of the
imagery. The movement and storage of items should be auditable and traceable.

Depending upon internal procedures, copies of imagery submitted for comparison may be
retained for a set period of time after completion of the examination, for example on a secure
internal network. This digital storage should equally be auditable and ideally cross referenced to
a database to ensure that data are managed and traceable at all times.

Consideration should be given to laboratory lighting conditions to optimise viewing of imagery

and minimise issues such as screen glare (e.g. using matte rather than glossy screens or
installing monitor hoods). Examiners should take regular screen breaks to prevent fatigue when
examining imagery.

When conducting independent verifications as part of the ACE-V process (see Section 5) the
position of monitors and desks should be considered. Ideally examiners should not be exposed
to imagery before conducting their structured ACE-V examination, in order to minimise possible
sources of bias, such as confirmation bias caused by viewing the reference imagery prior to the
questioned imagery. This can be challenging in open office environments but should be taken
into account when positioning desks and monitors and deciding the order in which examinations
are conducted and at what stage reference imagery is viewed.

4.5 Materials and Reagents

Questioned imagery submitted for FIC is typically uncontrolled and the quality of the imagery
will be beyond the examiners control. Questioned imagery will encompass a range of different
sources and quality.

The reference imagery used in the comparison should be captured in a controlled environment
and ideally include the following:
• Imagery captured and stored at a sufficient level of detail to include fine feature
detail such as scars, marks and small creases
• Imagery captured in a range of different camera angles and poses, including the
camera angle and pose of the questioned imagery
• Imagery captured at a similar time to the questioned imagery (if the questioned
imagery was captured a significant amount of time ago recent reference imagery of
the subject should also be supplied to highlight any age-related changes)

Reference imagery that does not meet the criteria above may be detrimental to the results of the

8/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

comparison and could cause errors [11]–[13]. If the reference imagery is unsuitable or of limited
scope for comparison the examiner should make efforts to source additional suitable reference
imagery.

Comparisons should not be restricted to single images (e.g. a still from a video). Where
possible, examiners should utilise a range of questioned and reference imagery when
conducting a comparison to provide the maximum amount of subject detail [14].

If possible, capturing additional reference imagery using the same or similar recording device
and conditions of the questioned imagery may be beneficial [15].

In some circumstances the examiner may be requested to compare multiple sources of

questioned imagery without reference imagery.

5. METHODS

5.1 Case Strategy

Prior to undertaking the FIC examination, a case strategy should be established defining the
following:
• The methods to be used in the examination
• Which imagery is to be compared
• Dates on which the imagery was taken
• Any additional relevant information concerning the subject or the imagery (e.g.
cosmetic surgery)
• Critical dates for completion of the examination
• Any other types of forensic examination and the order in which this will be conducted
(e.g. DNA profiling or fingerprint development)
• The propositions under evaluation

Ideally the case strategy should be set by a competent member of laboratory staff who does
not take part in the FIC examination, to mitigate contextual bias. The strategy setter can then
remove non-essential sources of information that may cause bias (e.g. identification by other
means, background information of the investigation) and retain relevant information that the
examiners need to know (e.g. dates of image capture).

The strategy setter should establish what question(s) or propositions are to be addressed by the
examination. The examination should not be aimed at addressing only one proposition, such as
the proposition that the subject in the CCTV is the defendant, as this may bias results towards
that outcome and other likely alternatives may not be considered. An approach to overcome
this is for at least two mutually exclusive propositions for the evaluation of the comparison
findings to be set as part of the case strategy for evaluation. Facial image comparison should
be reported at the source level rather than activity level. The propositions should reflect both the
view of the prosecution/mandating authority and the defence/alternative scenario. If a defence
statement has not been provided the examiner should set a competing proposition. When
setting the competing proposition the reporting examiner should adopt a scenario that likely and
reasonably reflects the party’s position, such as:

H1 - Subject A is the defendant

H2 - Subject A is another individual or Subject A is another adult male

9/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

The examiner should not adopt a competing proposition that is too broad as this will maximise
the evidential value of the observations. Likewise, the competing proposition should not be too
narrow as this may exclude realistic potential candidates. The propositions should be set prior
to commencing the examination. If new information comes to light during the examination that
alters the propositions this should be recorded and the strategy updated accordingly.

5.2 Technical Processes

5.2.1 Conversion of imagery

If the submitted imagery is supplied in a non-standard or proprietary format it may be necessary
to convert the imagery to another format for analysis or subsequent processing. Any processes
used to convert imagery should be validated to determine if there is any degradation to the
output. When converting imagery examiners should consider the following:
• The converted imagery is maintained at the native resolution of the input imagery
• The aspect ratio of the input imagery is maintained
• If transcoding to a new format the converted imagery is produced in a losslessly
compressed or uncompressed format
• Any colour space conversions do not cause a loss of facial detail or clipping of
colour values

Examiners should record the processes applied when converting the imagery, the software
used, the version number, parameters and format of the output file.

5.2.2 Digitisation of Analogue Imagery

The conversion of an analogue source to a digital file will inevitably result in the loss of some
information. The processes used to digitise analogue imagery should be validated to ensure that
there is not a significant loss of facial detail. When digitising analogue media examiners should
consider the following:
• Where possible adjustments to the imagery (e.g. brightness) should be applied in
the analogue domain before digitisation
• The digital file should be produced with lossless compression or in an
uncompressed format
• The resolution of the digital file is sufficient so that there is not a loss of facial detail
• The bit depth of the digital file is sufficient so that there is not a loss of facial detail
• The transmission of analogue signals should be via the highest quality pathway
available to minimise degradation of the signal

Examiners should record the make, model and serial number of digitisation hardware, any
software used and the version number, the format and parameters of the output file and any
processing applied.

5.2.3 Enhancement of Imagery

During the examination there may be the potential to enhance the submitted imagery
through processing, to produce a result that is better suited for the purpose of comparison
or provides additional clarity of facial feature detail. The ENFSI BPM for Forensic Image and
Video Enhancement (ENFSI-BPM-DI-02) [2] provides detailed guidance on how the forensic
enhancement of imagery should be conducted. This guidance also applies to imagery enhanced
for FIC. However not all of the processes mentioned within the ENFSI BPM for Forensic Image
and Video Enhancement (ENFSI-BPM-DI-02) should be applied in facial image comparison.
Image enhancement processes alter the appearance and content of an image and may distort

10/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

facial features or introduce artefacts that mislead the comparison examination. Any applied
enhancement procedures should be validated for their intended use.

Examiners should be competent in any enhancement processes applied to imagery and have
knowledge of what alterations to an image will occur from a process. Examiners should be
capable of identifying artefacts introduced by an enhancement process and any potentially
detrimental impacts on image quality. Additionally examiners should also consider the following
when attempting to enhance imagery for FIC:
• Digital enhancement processing should only be applied to a copy of the imagery
• The output format of the enhanced image should be losslessly compressed or
uncompressed
• Ideally the output format of the enhanced image should support the processing
steps of the enhancement to be revisited for transparency (e.g. using smart objects
or layers within .TIFF or .PSD files)
• A copy of the imagery without the enhancement should also be available for the
subjective analysis
• The introduction of artefacts into the image should be kept to a minimum
• Alterations to the facial appearance of the subject in the image should be kept to a
minimum
• Complex enhancement processing, such as image deconvolution, adaptive filtering
and inter-frame processing (e.g. super resolution) should be used with caution, due
to the high propensity of these techniques to introduce artefacts
• Any enhancement processes applied to an image for facial comparison should be
validated, insofar as possible, for its intended use

Examiners should record the type of processes used to enhance an image, the parameters
applied, the order in which processes were applied, the format and parameters of the output
file, the name and version number of the software used and a rationale for why a process was
applied.

If an enhanced image is to be used for FIC the examiner should always refer back to the original
un-enhanced image as well as the enhanced version when conducting their analysis, due to the
potential introduction of erroneous image artefacts by an enhancement process.

5.3 Subjective Examination

The ENFSI Digital Imaging Working Group (DIWG) and FISWG currently recommend that
morphological analysis be used for FIC. Empirical evidence indicates that this method improves
accuracy in facial comparison tasks [16]. Morphological analysis is defined as follows [17]:

‘Morphological analysis as a comparison method is based on the assessment of

correspondence of the shape, appearance, presence and/or location of facial features.
These features include global (corresponding to the overall face), local (including
anatomical structures such as nose or mouth and their components, e.g., nose bridge,
nostrils, ear lobes) and discriminating characteristic facial marks such as scars or moles.’

It is recommended that a facial feature checklist with standardised terminology is used by

a laboratory, such as the FISWG facial feature list [18]. This will provide consistency and
repeatability between different examiners in the laboratory when using the morphological
approach. Table 1 shows an example of the facial features used by the FISWG list.

11/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Table 1 - Facial components of the FISWG Facial Feature List

ID Facial Components
1 Skin
2 Face/Head Outline
3 Face/Head Composition
4 Hairline/Baldness Pattern
5 Forehead
6 Eyebrows
7 Eyes
8 Cheeks
9 Nose
10 Ears
11 Mouth
12 Chin/Jawline
13 Neck ID Facial Components

14 Facial Hair 1
2
Skin
Face/Head Outline

15 Facial Lines 3
4
Face/Head Composition
Hairline/Baldness Pattern
16 Scars 5
6
Forehead
Eyebrows
17 Facial Marks 7
8
Eyes
Cheeks
18 Alterations 9 Nose
10 Ears
19 Other 11 Mouth
12 Chin/Jawline
13 Neck
14 Facial Hair
General facial features (aka class feature detail) should be further sub-divided into sub-features
15 Facial Lines

9-Noseof higher quality imagery. Figure 1 shows an

(aka sub-class feature detail) to allow the analysis
16
17
Scars
Facial Marks
Component
example Characteristics
of sub-class Characteristic Descriptors
feature detail for the nose.
18
19
Alterations
Other
9.1 Nasal Outline (Profile and Front view) • Overall Shape
• Length and/or width
Component Characteristics
9-Nose
Characteristic Descriptors
• Prominence
9.1 Nasal Outline (Profile and Front view) • Overall Shape
• Length and/or width
• Symmetry • Prominence
• Symmetry
9.2 Nasal Root (Bridge) • RootFront
9.2 Nasal (Bridge) View: width,• length, Front View:shape, depth
width, length, shape, depth
• Profile View: length, depth, angle
• BodyProfile View: length,
9.3 Nasal depth,
• Front angle
View: width, length, shape, angle
• Profile View: length, angle, contour
9.3 Nasal Body • Tip Front View: width,• length,
9.4 Nasal shape,
Shape (in front angle
and profile view)
• Angle (e.g. up, down)
• Profile View: length, angle, contour
• Symmetry
9.5 Nasal Base
9.4 Nasal Tip • Shape (in front and• profile
• Width
Height view)
• Base:Angle
9.6 Nasal Alae (Wings(e.g. up, down)
of Nose)
• Deviation to the right or left
• Thickness
• Symmetry • Symmetry
• Shape
9.5 Nasal Base • Width
9.7 Nasal Base: Nostrils (Nasal Openings) • Shape and size of opening
• Symmetry
• Height
9.8 Nasal Base: Columella (Soft Tissue
• Hair
• Width and length
between Deviation to the right
• Nostrils) or left
• Relative position
• Symmetry
9.6 Nasal Base: Alae (Wings of Nose) • Thickness
Figure 1 – Sub-class characteristics of the nose from the FISWG facial feature list
• Symmetry
• Shape
In addition to facial features the checklist should also include fine feature detail, such as scars,
9.7 Nasal
marks Base: creases
and small Nostrils (Nasal Openings)such as •piercings
and alterations Shapeand andtattoos.
size of opening
Fine feature detail can
• Symmetry
• Hair
12/50
9.8 Nasal Base: Columella (Soft Tissue • Width and length
between Nostrils) • Relative position
• Symmetry
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

prove very useful in FIC when used in combination with facial feature detail, and can distinguish
between identical twins [19].

The following comparison methods are currently not recommend for FIC of uncontrolled
imagery. Current empirical research indicates that these methods may not produce accurate
comparison results and have issues concerning repeatability and reliability:
• Facial feature classification [20], [21]
• Photo anthropometry/proportional alignment [22], [23]
• Superimposition/overlaying [24], [25]

A brief review of the above facial comparison techniques is provided in Appendix C.

In the absence of empirical research supporting the use of a method in casework,

any method used to analyse and compare facial imagery should be validated by the
laboratory to demonstrate the accuracy of the method and show that it produces reliable
results. The laboratory should also state why the method has been used, supported by
the results from any relevant validation studies.

In order to perform the examination and comparison of facial imagery an ACE-V workflow is
recommended. This approach is characterised by applying the following subjective phases of
the examination: analysis, comparison, evaluation and verification. Although ACE-V is described
as a linear workflow examiners may not necessarily work completely in a linear fashion.

5.3.1 Analysis
The analysis of the questioned imagery is the first stage of the process. The analysis phase
is intended to determine whether the quality of the questioned imagery is sufficient for
meaningful facial comparison. The analysis phase consists of an assessment of the quality of
the questioned imagery to determine the degree of facial detail present and what features are
available for comparison. The reference imagery should not be analysed in detail at this stage
to mitigate the risk of confirmation bias (i.e. being potentially misled by viewing features in high
quality reference imagery that are not resolvable in the lower quality questioned imagery).

Imagery of low quality may result in limited or at worst unreliable findings in FIC [26]. The
image quality assessment of the questioned imagery should address, but not be limited to, the
following aspects:
• Pixel resolution of the face as determined by the overall resolution of the image and
the distance of the subject from the camera
• Compression artefacts visible within the imagery causing a irrecoverable loss of
detail
• The lighting conditions of the image and the impact this has on the visibility of
features and the degree of contrast between different features
• Blur within the image causing a potentially irrecoverable loss of detail
• Geometric distortions such as lens barrelling that may distort the appearance of
facial features
• If the angle of the camera is causing occlusion of facial features
• If other factors such as clothing or physical obstruction are occluding facial features

During analysis the examiner should note if there are factors affecting the appearance of the
observed features, such as expression, pose, clothing or image quality issues (see Appendix B
for details).

13/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

When analysing the questioned imagery all the relevant submitted material should be reviewed.
Specific images may then be selected that provide the most facial detail for comparison and at
a similar pose and camera angle to the reference imagery. Where possible, observations should
not be restricted to just a single image. Features only observed in a single image, particularly if
that image is of low quality, may be attributable to image artefacts.

Examiners should also ascertain whether the imagery has been submitted in its original format,
as generally this will provide the highest quality imagery for examination. Examples of converted
imagery that may have introduced a loss of quality include:
• Screen grabs of video
• Transcoded video (e.g. conversion to AVI or DVD video formats)
• Still images taken from a video
• Photographs of a display screen replaying imagery
• Still images resaved in a lossy format (e.g. JPEG)
• Digital conversions of analogue imagery
Examiners should be competent in establishing if imagery is presented in its original format.
If imagery is not provided in its original format and the examiner believes this has introduced
a loss in quality from the original or may have significantly altered the imagery the examiner
should request the original imagery from the investigation or an explanation of why the original
is not available.

If the findings of the analysis phase are that there is insufficient facial feature detail for reliable
comparison this decision should be reviewed by another competent examiner. If the reviewer is
in agreement with the findings the party requesting the comparison should be contacted as soon
as practicable to inform them of the result.

5.3.2 Comparison
The aim of this stage is to systematically compare the facial features observed in the questioned
and reference imagery to establish any similarities or differences between the observed
features. The comparison should follow the morphological approach and utilise a standard
reference facial feature list [18].

When making observations the examiner should consider the impact of image quality factors on
apparent similarities and differences, such as resolution, occlusion or motion blur (see Appendix
B for further details). For example, if a difference or similarity is observed but it is readily
explainable by imaging factors this should be noted as it will impact upon the evaluation stage.

When conducting the comparison ideally the questioned and reference imagery should:
• Be captured on a similar date
• Be captured at a similar camera angle
• Show the subject in a similar pose and expression
• Be captured under similar lighting conditions

If the above are not met this could impact upon the reliability of the observations and limit the
suitability of the imagery for meaningful comparison. Empirical evidence has demonstrated
reduced accuracy in facial comparison tasks when these conditions are not met [13], [27].

During the comparison the examiner may resize or rotate the imagery to aid in the observation
of similarities and differences. Examiners should have an awareness of the impact of different

14/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

image resizing algorithms on the appearance of facial feature detail. When resizing any image
for FIC the aspect ratio of the image must be maintained.

Rotating an image will also result in the interpolation of pixels and alteration to the image. When
rotating images for comparison the examiner should rotate the highest quality image (typically
the reference image) as this will have less of an impact on image detail than rotating a lower
quality image.

5.3.3 Evaluation
See section 12

5.3.4 Verification
The FIC ACE process should be repeated by a second independent examiner to provide a
verification of the result.

The term verification may be used to refer to the following processes [28]:
• Blind verification: an independent verification done by a second examiner using the
ACE method, without knowing the conclusion of the first examiner
• Non-blind verification: an independent verification done by a second examiner using
the ACE method, knowing the conclusion of the first examiner

Due to the subjective nature of FIC a blind verification is recommended for all examinations
as this will help prevent confirmation bias that can occur in non-blind verification. Agency
SOPs should define how the stages of the ACE processes are repeated during the verification.
Verification should be performed on all examinations that progress to the evaluation stage.

The verification should be documented to the same level of detail as the initial examiner’s
examination.

Where there are sufficient resources there should also be a critical findings check of the
conclusions as part of the peer-review process (see Section 5.3.6).

5.3.5 Resolving Disagreements

Agencies should have written procedures to address differences in interpretation by different
examiners arising from the verification. As FIC is a subjective process it is expected that
there will be differences in how features and aspects of an image are described by different
examiners, but these differences may not have a significant impact upon the overall findings.

The resolution procedure should define when differences in interpretation are treated as
significant, supported by validation of the process and define steps to resolve the difference in
interpretation (e.g. a third independent verification or a panel review process).

5.3.6 Peer Review

The findings of the FIC examination should be peer reviewed before release to the customer. The
peer review should comprise of two aspects; a critical findings check and a technical fact check.
• Critical findings check: A competent facial examiner reviews the ACE-V process
of the examination and checks the findings of the examination to ensure they are
consistent with the documented process.
• Fact check: The technical aspects of the examination are checked for consistency,
such as dates of when an examination was conducted or the seal numbers of exhibit
continuity bags.

15/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Errors found during the fact check should be corrected by the examiner responsible for reporting
the conclusion before release to the customer. Errors encountered during the critical findings
check should be addressed using the agencies policy for resolving disagreements (Section 5.3.5).

6. VALIDATION AND ESTIMATING UNCERTAINTY OF MEASUREMENT

6.1 Validation
The laboratory should use validated methods for the processing, examination and evaluation of
facial imagery. Before a validation process can be started the critical aspects of the method or
procedure should be identified and the limitations defined. This information should be recorded
in a validation plan as per laboratory SOPs.

During validation it is important to investigate the following aspects of the method:

• Accuracy
• Precision
• Range
• Repeatability
• Reproducibility
• Robustness

The validation process must be developed in accordance with the provisions of section 5.4.5 of
ISO 17025 [29]. At the time of writing there are no published standards addressing procedures
used in the field of FIC.

The final aim of the validation process is to establish whether the method meets the required
accuracy, precision, repeatability, reproducibility and robustness of its intended use. The results
of the validation study should be produced in a report as per laboratory SOPs.

Additional guidance of validating laboratory procedures can be found in the ENFSI Guideline
for the Single Laboratory Validation of Instrumental and Human Based Methods in Forensic
Science [30]

6.2 Estimating Uncertainty of Measurement

At present, it is not possible to calculate the probability of a match between two faces. This is
due to the absence of relevant population data, i.e. the frequency of occurrence of different
facial features within the relevant. In the absence of such data an estimation of the error in FIC
should be established through black box studies, validation studies, proficiency testing and
collaborative exercises.

Laboratories should attempt to identify the potential sources of uncertainty during examinations, as
stipulated in paragraph 5.4.6 of the ISO 17025 standard. Potential sources of uncertainty examiners
may encounter during the examination of materials includes, but is not limited to, the following:

• Sample size: the quantity of imagery available for analysis and comparison may
impact upon the strength of the evaluation. If the observations made during the
examination are limited to a single frame there may be less information available
compared to using multiple images or moving video.
• Quality of imagery: Image quality depends on a myriad of inter-related factors
including resolution, compression and occlusion of features. These factors will have
an intrinsic impact on the results of any examination. See Appendix B for examples
of factors affecting facial appearance.

16/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

• Human error: At each stage of the examination the actions of the human examiner
are critical to result. Human error may introduce uncertainty and should be
countered using processes such as the following:
• Structured training
• Ongoing competency testing
• Adhering to standard operating procedures
• Using validation methods
• Structured examinations (e.g. ACE-V)
• Independent verification of findings
• Fact checking and critical finding checks (peer-review)
• Participation in collaborative exercises and proficiency testing

7. PROFICIENCY TESTING

Proficiency tests should be used to test and assure the reliability of FIC processes. ENFSI
Guidance on the conduct of Proficiency Tests and Collaborative Exercises [9] provides
information on the organisation and frequency of proficiency tests. ENFSI also maintains a list
of proficiency test and collaborative exercise providers that may be consulted. When developing
proficiency tests providers should be aware of relevant scientific literature concerning the testing
of human performance [10].

ISO 17025 accreditation requires laboratories to perform at least one proficiency test per year [29].

8. HANDLING ITEMS

Media containing digital data should be handled in accordance with the ENFSI BPM for
Forensic Information Technology [31]. Analogue media, such as VHS cassettes, should be
stored in a safe and secure environment to prevent degradation of the material. Submissions
requiring more than one type of work should follow laboratory SOPs for dual submissions.
If the type of work requested will be impacted by the facial comparison examination (e.g.
DNA profiling, fingerprint enhancement) the examiner should consult with the investigator to
determine which examination should take place first.

9. INITIAL ASSESSMENT

Prior to the formal submission of imagery for FIC the laboratory may consider providing an initial
triage service to assess the suitability of the imagery and the proportionality of the request. For
example, if the quality of the imagery is likely to only provide weak or inconclusive results or
there is substantial other evidence in the investigation FIC may not be appropriate. It is likely
at this stage only a sample of the imagery will be triaged for quality assessment. If the request
is proportionate and the sample of imagery sufficient for examination a full examination should
be conducted. Triaging cases can be an effective way of managing caseloads and ensuring
operational effectiveness. It also manages the expectations of the investigator regarding how
beneficial facial comparison will be to their case.

17/50
 BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

745 10. PRIORITISATION AND SEQUENCE OF EXAMINATIONS

746 10. PRIORITISATION AND SEQUENCE OF EXAMINATIONS
747 A FIC examination can be broadly separated into technical processes and subjective
748
A FIC examination
examination [32]: can be broadly separated into technical processes and subjective examination [32]:
749
750
Technical Processes Conversion of
751
imagery
752
753
754 Processing of
755 imagery
756
757
758 Subjective Examination Analysis of
759 questioned imagery
760
761
762 Comparison of
questioned and
763
reference imagery
764
765
766 Evaluation of
767 examination findings
768
769
770 Independent
verification
771
772
773 Figureof2the
Figure 2 – Overview – Overview of the FIC
FIC examination examination process
process
774
775 Depending upon
Depending upon laboratory
laboratory organisation,
organisation,thethe
facial examiner
facial may may
examiner undertake the technical
undertake the
776 processing of imagery or this may be undertaken by a specialist in image
technical processing of imagery or this may be undertaken by a specialist in image processing then
777 passed onto the facial examiner. In either arrangement it is necessary to ensure
processing then passed onto the facial examiner. In either arrangement it is necessary that staff
undertaking any task are competent to do so and have sufficient training and knowledge in that
778 to ensure that staff undertaking any task are competent to do so and have sufficient
field. A clear chain of continuity must be maintained throughout all stages of the examination,
779 training and knowledge in that field. A clear chain of continuity must be maintained
particularly if material is moving between different departments.
780 throughout all stages of the examination, particularly if material is moving between
781 different departments.
The subjective examination of FIC should be undertaken within a structured ACE-V workflow.
782 Although ACE-V is described as a linear workflow examiners should not necessarily work
783 The subjective
completely in aexamination
linear fashion, of e.g.
FIC processing
should be ofundertaken
the imagerywithin
may aoccur
structured ACE-V
during any stage of the
784 subjectiveAlthough
workflow. examination
ACE-Vor a isfeature withinas
described theaquestioned imagery
linear workflow may onlyshould
examiners be observed
not during
785 the comparison
necessarily stage. This in
work completely is aacceptable as long
linear fashion, e.g.as the examiner
processing of thecan justifymay
imagery theiroccur
deviation from
786 a linear
during workflow.
any stage of the subjective examination or a feature within the questioned imagery
787 may only be observed during the comparison stage. This is acceptable as long as the
788 examiner can justify their deviation from a linear workflow.
789
790
791

Ref code: ENFSI-BPM-DI-01 Issue No. 001 Page: 18 / 59

18/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

11. RECONSTRUCTION OF EVENTS

When sourcing reference material of a subject for FIC, in addition to high quality reference im-
agery, attempts should be made to recreate, insofar as possible, the capture conditions of the
questioned imagery, as discussed in Section 4.5.

12. EVALUATION AND INTERPRETATION

The ENFSI Guideline for Evaluative Reporting in Forensic Science [3] provides forensic
examiners with a framework for formulating evaluative reports and contains an example of
evidence evaluation from a FIC examination. This guideline should be consulted for specific
guidance on formulating logical, evaluative opinions. This BPM provides details of some of the
factors that can influence evaluation in FIC and should be read in conjunction with the ENFSI
Guideline.

During the evaluation stage the observed similarities and differences from the comparison are
evaluated by the examiner, resulting in a conclusion that states the evidential weight as a level
of support for one of the competing propositions. FIC is a subjective process and currently it
is not possible to assign a quantitative probability to the examination findings. Therefore the
conclusions will be based upon the training, knowledge and experience of the examiner.

The findings should be expressed as a level of support for one of the propositions, for example:
• The findings from the examination support the view that the person indicated on
the CCTV still is the person in the reference images (level +2), rather than that the
individual is another adult male [3].
• The examination findings provide strong support for the proposition that Subject A is
another individual, rather than the defendant.

The propositions under consideration will depend on the circumstances of the case and should
be established prior to undertaking the FIC examination.

Support levels are typically reported using a graded scale. At present there is not a universally
accepted scale for reporting FIC conclusions and there is a wide range in scales used by
different agencies. Whichever way the conclusions of a FIC examination are reported should
be validated to ensure that it is applicable to the type of imagery encountered in casework and
produces repeatable, reliable and accurate results.

Due to the lack of empirical data regarding the frequency of occurrence of facial features within
the population, it is recommended that statements of absolute certainty are not used when
reporting FIC conclusions.

If using a graded scale to report conclusions the following factors should be considered:
• The scale should include degrees of support for identity correspondence and
degrees of support for exclusion
• The scale should contain an inconclusive level (0) in which the facial examiner can
express their opinion as offering no support to either proposition
• The number of support levels used in the scale should be sufficient to provide a
meaningful range of opinions whilst still ensuring repeatability between different
examiners. This should be established through validation studies and competency testing

19/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Evaluating the strength of the examination findings is complex and dependant on a number of
interacting factors, such as:
• The extent to which facial features can be observed in the imagery
• The number of facial features visible within the imagery
• How common the observed facial features are (this will primarily be based upon
subjective knowledge and experience)
• The transience/permanence of the observed facial features
• The quantity and nature of the similarities or differences observed
• If the similarities or differences are readily explainable as resulting from image
quality factors
• The correspondence in imaging conditions between the questioned and reference
imagery (e.g. resolution, lighting, camera angle, occlusion)
• The difference in time between the capture of the questioned imagery and the
capture of the reference imagery

Each of these factors should be addressed as a condition that impacts the level of support
that can be offered for one of the propositions. For example, if only a small number of poorly
resolved facial features are observed to be similar and the images are captured at non-matching
angles the findings may only offer limited or weak support to the proposition the images depict
the same person. If there are differences observed in the detailed features of the face and in
some fine feature detail (such as facial marks), the images have been captured at a similar time
and under similar conditions then the findings may offer the strongest levels of support for the
proposition that the images depict different individuals.

Examiners should be aware of issues concerning the interpretation of conclusion scales by lay
audiences reported in the literature [33].

Examiners conducting the evaluation stage should be trained and competent to do so with
specialised knowledge relevant to this task.

13. PRESENTATION OF EVIDENCE

The overriding duty of those providing expert testimony is to the court and to the administration
of justice. As such, evidence should be provided with honesty, integrity, objectivity and
impartiality.

Evidence can be presented to the court either orally or in writing. Only information which is
supported by the examinations carried out should be presented. Presentation of evidence
should clearly state the results of any evaluation and interpretation of the examination. Written
FIC reports should include all the relevant information in a clear, concise, structured and
unambiguous manner as required by the relevant legal process. Written reports must be peer
reviewed.

Expert witnesses should resist responding to questions that take them outside their field
of expertise unless specifically directed by the court, and even then a declaration as to the
limitations of their expertise should be made.

20/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

If an evaluative conclusion is made based on relevant population data the source of the data
should be made clear in the written or oral evidence. If the evaluative opinion is based upon
subjective knowledge, training and experience this should be stated also. Further details on
reporting evaluative conclusions in FIC can be found in the ENFSI Guideline for Evaluative
Reporting in Forensic Science.

14. HEALTH AND SAFETY

Which health and safety considerations are relevant to be considered regarding FIC examina-
tions will depend upon the operational requirements of the agency or organisation for whom
the examiner works and the types of casework undertaken. Laboratory SOPs should address
working with hazardous material and/or material of an indecent or disturbing nature if applicable.
When viewing imagery for prolonged periods examiners should take regular screen breaks.

15. REFERENCES

[1] ASTM E2916 Terminology for Digital and Multimedia Evidence Examination.
[2] European Network of Forensic Science Institutes, “DRAFT Best Practice Manual
for Forensic Image and Video Enhancement,” (ENFSI-BPM-DI-02) 2018
[3] European Network of Forensic Science Institutes, “ENFSI guideline for evaluative
reporting in forensic science: Strengthening the Evaluation of Forensic Results
across Europe,” 2015.
[4] Facial Identification Scientific Working Group, “Guidelines and Recommendations
for Facial Comparison Training to Competency,” Facial Identification Scientific
Working Group, 2010.
[5] A. M. Burton, D. White, and A. McNeill, “The Glasgow Face Matching Test.,”
Behav. Res. Methods, vol. 42, no. 1, pp. 286–291, 2010.
[6] A. J. Dowsett and A. M. Burton, “Unfamiliar face matching: Pairs out-perform
individuals and provide a route to training,” Br. J. Psychol., vol. 106, no. 3, pp.
433–445, 2015.
[7] D. White, R. I. Kemp, R. Jenkins, and A. M. Burton, “Feedback training for facial
image comparison,” Psychon. Bull. {&} Rev., vol. 21, no. 1, pp. 100–106, 2014.
[8] Forensic Science Regulator, “Codes of Practice and Conduct - Appendix
FSR-C-107 - Digital Forensic Services,” no. 1, pp. 1–49, 2014.
[9] European Network of Forensic Science Institutes, “Guidence on the conduct of
proficiency tests and collaborative exercises within ENFSI,” no. 2. pp. 1–9, 2012.
[10] K. A. Martire and R. I. Kemp, “Considerations when designing human
performance tests in the forensic sciences,” Aust. J. Forensic Sci., pp. 1–17, Nov.
2016.
[11] M. Bindemann, J. Attard, A. Leach, and R. A. Johnston, “The effect of image
pixelation on unfamiliar-face matching,” Appl. Cogn. Psychol., vol. 27, no. 6, pp.
707–717, 2013.
[12] H. U. Keval and M. A. Sasse, “Can we ID from CCTV: Image quality in digital
CCTV and face identification performance,” Proc. SPIE, vol. 6982, p. 69820K,
2008.
[13] A. M. Megreya, A. Sandford, and A. M. Burton, “Matching face images raken on
the same day or months apart: The limitations of photo ID,” Appl. Cogn. Psychol.,
vol. 27, no. 6, pp. 700–706, 2013.

21/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

[14] D. White, A. M. Burton, R. Jenkins, and R. I. Kemp, “Redesigning photo-ID to

improve unfamiliar face matching performance.,” J. Exp. Psychol. Appl., vol. 20,
no. 2, pp. 166–73, 2014.
[15] A. M. Burton, P. Miller, V. Bruce, P. J. B. Hancock, and Z. Henderson, “Human
and automatic face recognition: A comparison across image formats,” Vision
Res., vol. 41, no. 24, pp. 3185–3195, 2001.
[16] A. Towler, D. White, R. I. Kemp, A. Towler, D. White, and R. I. Kemp, “Evaluating
the Feature Comparison Strategy for Forensic Face Identification,” J. Exp.
Psychol. Appl., 2017.
[17] Facial Identification Scientific Working Group, “Guidelines for Facial Comparison
Methods.” 2012.
[18] Facial Identification Scientific Working Group, “Facial Image Comparison Feature
List for Morphological Analysis.” 2013.
[19] S. Biswas, K. W. Bowyer, and P. J. Flynn, “A study of face recognition of identical
twins by humans,” 2011 IEEE Int. Work. Inf. Forensics Secur. WIFS 2011, no.
November, 2011.
[20] A. Towler, D. White, and R. I. Kemp, “Evaluating training methods for facial image
comparison: The face shape strategy does not work,” Perception, vol. 43, no.
2–3, pp. 214–218, 2014.
[21] S. Ritz-Timme et al., “A new atlas for the evaluation of facial features:
Advantages, limits, and applicability,” Int. J. Legal Med., vol. 125, no. 2, pp. 301–
306, 2011.
[22] R. Moreton and J. Morley, “Investigation into the use of photoanthropometry in
facial image comparison,” Forensic Sci. Int., vol. 212, no. 1–3, pp. 231–237,
2011.
[23] K. F. Kleinberg and P. Vanezis, “Variation in proportion indices and angles
between selected facial landmarks with rotation in the Frankfort plane.,” Med. Sci.
Law, vol. 47, no. 2, pp. 107–16, 2007.
[24] A. Strathie, A. Mcneill, and D. White, “In the Dock: Chimeric Image Composites
Reduce Identification Accuracy,” Appl. Cogn. Psychol., vol. 26, no. 1, pp. 140–
148, 2012.
[25] A. Strathie and A. McNeill, “Facial Wipes don’t Wash: Facial Image Comparison
by Video Superimposition Reduces the Accuracy of Face Matching Decisions,”
Appl. Cogn. Psychol., vol. 30, no. 4, pp. 504–513, 2016.
[26] K. Norell, K. B. Läthén, P. Bergström, A. Rice, V. Natu, and A. O’Toole, “The
Effect of Image Quality and Forensic Expertise in Facial Image Comparisons,” J.
Forensic Sci., vol. 60, no. 2, pp. 331–340, 2015.
[27] R. Jenkins, D. White, X. Van Montfort, and A. Mike Burton, “Variability in photos
of the same face,” Cognition, vol. 121, no. 3, pp. 313–323, 2011.
[28] European Network of Forensic Science Institutes, “Best Practice Manual for
Fingerprint Examination ENFSI-BPM-FIN-01,” vol. 1, no. November, pp. 1–60,
2015.
[29] EN ISO/IEC 17025:2005, “General requirements for the competence of testing
and calibration laboraties.” .
[30] European Network of Forensic Science Institutes, “Guidelines for the single
laboratory Validation of Instrumental and Human Based Methods in Forensic
Science,” no. 1, pp. 1–31, 2014.
[31] European Network of Forensic Science Institutes, “ENFSI Best Practice Manual
for the Forensic Examination of Digital Technology ENFSI-BPM-FIT-01, 2015.

22/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

[32] Metropolitan Police Service, “Forensic Image Comparison and Interpretation

Evidence : Guidance for Prosecutors and Investigators Issue 2,” no. 2. 2016.
[33] D. McQuiston-Surrett and M. J. Saks, “The testimony of forensic identification
science: What expert witnesses say and what factfinders hear,” Law Hum.
Behav., vol. 33, no. 5, pp. 436–453, 2009.

16. AMENDMENTS AGAINST PREVIOUS VERSION

Not applicable (first version).

23/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

APPENDIX A: EDUCATION AND TRAINING GUIDELINES FOR FACIAL EXAMINERS

The purpose of this guideline is to provide a training outline for persons involved in delivering
facial comparison training courses, and to assist in developing a uniform formal training program
for facial image examiners.

This training guideline is specific to facial examiners conducting facial comparison

examinations, defined as follows:

‘The task of facial examination includes, but is not limited to, a rigorous one-to-
one analysis, comparison, and evaluation of controlled and uncontrolled images
for the purpose of effecting a conclusion. Examiners in this situation have to draw
on a larger foundation of knowledge, skill, and ability to accurately reach their
conclusions. Additionally, the articulation of the scientific and legal basis for the
expression of conclusions for many forensic, intelligence, or law enforcement
purposes requires an even more advanced level of training to include an expanded
set of knowledge, skills, and abilities above the level of basic concepts.’1

Facial comparison is an innate ability and performance varies significantly across different
individuals2. The laboratory should implement a facial comparison testing programme to gauge
examiner ability before training. There is a wide range in innate ability when comparing unfamiliar
faces, such a testing scheme can be used to highlight individuals with high innate performance
for recruitment as facial examiners and enrolment in subsequent training and mentoring.

Throughout training regular testing and evaluation of trainee performance should be conducted
using ground truth facial comparison tasks. These tasks should cover the range of imagery that
will typically be encountered in casework. Empirical research has demonstrated that providing
trainees with feedback on facial comparison tests improves trainee ability3.

IMPORTANT NOTE

This guide is designed to provide a comprehensive list of training topics relevant to

forensic facial comparison. Not all topics will be relevant to every agency or organisation
and not all topics are required to become a competent facial examiner.

Which topics are required to achieve competency will depend upon the operational
requirements of the agency or organisation and the types of casework undertaken.

Throughout this document the terms awareness, knowledge and ability are used in the context
of the following definitions:

• Awareness: The trainee should have a general acquaintance with the relevant
major elements of a given method or technology to include specific capabilities and
limitations.

1
Guidelines and Recommendations for Facial Comparison Training to Competency (Section 3),
FISWG, Version 1.1
2
White et al. 2010, The Glasgow Face Matching Test, Behavior Research Methods, 42 (1), 286-291
3
White et al. 2014, Feedback training for facial image comparison, Psychon Bull Rev, 21 (1), 100-106

24/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

• Knowledge: The trainee should have an in-depth knowledge of the topic and how the
subject relates to their discipline, acquired through education, training and experience.
• Ability: The trainee should have the necessary capability to carry out a particular
skill or process, acquired through training, proficiency testing and experience. The
trainee should also be aware of the limitations of the process.

Facial comparison training programmes should aim to provide knowledge and ability in
the following:
Knowledge of:
• Principals of image comparison
• Psychology of image comparison (e.g. innate ability vs. training)
• Principals of digital image and video capture and recording and effect on quality
(including resolution, compression, interlacing etc.)
• Principals of proprietary video (e.g. CCTV)
• Different illumination in video and images
• Analogue media
• Principals of image processing using global and local filters and their limitations and
pitfalls
• ACE-V process: Analyse – Compare – Evaluate – Verify
• Methods of comparison and their limitations
• Cognitive Bias
• Methods of Evaluation
• Facial anatomy and the commonality of facial features
• Age-related and environmental changes to facial features
• Forensic Reporting
• Legislation
• Procedures applied in the laboratory and in the scene of crime, including chain-of-
custody, as well as procedures related to law (court testimony)
• Relevant health and safety considerations (e.g. working with hazardous material or
material of an indecent or disturbing nature)
• Quality management systems and working practices at the laboratory
Ability in:
• Extracting still images from video and maintaining image quality
• Digitising analogue media and minimising image quality loss
• Transcoding images without degradation in quality
• Identifying original recordings
• Replaying proprietary video (e.g. CCTV)
• Using enhancement techniques in images and video
• Image resizing, cropping and rotation
• Assessing imagery for comparison
• Following the appropriate method for comparison and documenting that process
• Describing and comparing facial features observed in an image
• Determining age related and environmental differences between images
• Determining the likely cause of similarities and differences observed between two
images
• Determining the commonality of facial features observed in an image and evaluating
the significance of observations, using either population data or subjective knowledge
• Correct interpretation of the results obtained
• Case management, reporting, communication of analytical findings, presentation
and defending them in court
• Peer-review of critical findings

25/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

1. OVERVIEW

1.1 Fundamentals of Facial Image Comparison

1.1.1. Objectives
• Knowledge of the principles of comparison
• Knowledge of standard terminology
• Awareness of the limitations of image comparison as a method of identification
• Knowledge of the different processes of facial identification (i.e. facial image
comparison, automated facial recognition, human cognitive facial recognition)
• Knowledge of the psychology of facial identification (e.g. innate ability vs. trained
ability and own-race effect)

1.1.2. Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Review of example cases
• Clarification on questions
• Discussion

1.1.3. References
• “Facial Comparison Overview (Section 2)” FISWG
• “Forensic Image Comparison and Interpretation Evidence: Guidance for Prosecutors
and Investigators” GOV.UK
• “Guidelines for Facial Image Comparison Methods (Section 6)” FISWG
• “Best Practices for Forensic Photographic Comparison (Section 16)” SWGIT
• “Best Practices for Forensic Video Analysis (Section 7)” SWGIT
• “Digital and Multimedia Evidence Glossary” SWGIT/SWGDE
• “Glossary (Section 1)” FISWG
• “Perceptual expertise in forensic facial image comparison” White, D. Phillips, P. J.
Hahn, C. A. Hill, M. O’Toole, A. J. The Royal Society Proceedings B, Vol. 282 (1814)
• “Law’s Looking Glass: Expert Identification Evidence Derived from Photographic
and Video Images” Edmond, G. Biber, K. Kemp, R. I. Porter, G. Current Issues In
Criminal Justice, Vol. 20 (3), 2009
• “Stockwell Revisited: The Unhappy State of Facial Mapping” Campbell-Tiech and
Brynes 6 Archbold News 4, 2005
• “Limitations in Facial Identification: The Evidence”, Jenkins, R. Burton, A. M. Justice
of the Peace , 172, 5-12, 2008
• “The Glasgow Face Matching Test”, Burton, A. M. White, D. McNeill, A. A. Behavior
Research Methods, 42 (1), 286-291, 2010
• “The other-race effect does not rely on memory: evidence from a matching task”
Megreya, A. M. White, D. Burton, A. M. The Quarterly Journal of Experimental
Psychology, 64:8, 1473-1483

1.1.4. Assessment
• Quick decision facial comparison test with normalised test data (to determine innate
ability)
• Study questions (oral, written)
• Critical review of case examples

26/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

2. IMAGING FACTORS

2.1 Image capture and recording

2.1.1. Objectives
• Awareness of interlacing
• Awareness of proprietary file formats (e.g. digital CCTV)
• Awareness of common types of digital image and video compression algorithms
(e.g. JPEG, H.264)
• Awareness of different types of noise and its impact on image quality

• Knowledge of digital image capture

• Knowledge of the types, causes and impact of lens distortion
• Knowledge of the impact of scene illumination by different light sources (including
near infrared) and camera white balance settings on imagery
• Knowledge of common image resolutions
• Awareness of different image interpolation algorithms and their impact on image
quality
• Knowledge of the relationship between storage aspect ratio (SAR), pixel aspect
(PAR) and display aspect ratio (DAR) and their impact on image appearance
• Knowledge of the impact of compression on image quality

• Ability to determine if a file is an original recording

• Ability to extract still images from video and maintain quality
• Ability to transcode digital video and maintain quality
• Ability to determine if a video is interlaced or progressive

2.1.2. Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Practical demonstrations

2.1.3. References
• “Fatal Flaws: Uncertainty in the Interpretation of Colour in CCTV Images”, L. W.
MacDonald, Annals of the BMVA Vol. 2007, No. 7, pp 1−11 2007
• Colour analysis and verification of CCTV images under different lighting conditions,
R. A. Smith, K. MacLennan-Brown, J. F. Tighe, N. Cohen, S. Triantaphillidou, L. W.
MacDonald Imaging Technology Research Group, University of Westminster 2008
• Combining visible and near infrared images for realistic skin smoothing,
Fredembach, C. Barbuscia, N. Susstrunk, S. School of Computer and
Communication Science, Ecole Polytechnique Federale de Lausanne (EPFL),
Lausanne, Switzerland
• “Best Practices for Forensic Video Analysis (Section 16)” SWGIT
• “Issues Relating to Digital Image Compression and File Formats (Section 19)”
SWGIT
• “Digital Imaging Procedure” Home Office Scientific Development Branch
• The Principles and Practice of CCTV 3rd Ed. - Online version
• Cambridge in Colour Tutorials

27/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

2.1.4. Assessment
• Study questions (oral, written)
• Practical test

2.2 Analogue Media

2.2.1. Objectives
• Awareness of common types of analogue media
• Awareness of different printing techniques and how these affect the choice of
scanning procedure and filtering
• Awareness of the effect of different hardware on replay quality
• Knowledge of the correct storage and handling of analogue media to prevent
degradation
• Ability to digitise analogue media and minimise image quality loss
• Ability to scan printed documents at sufficient resolution and lossless compression
to minimise image quality loss
• Ability to apply image adjustments in the analogue domain prior to digitisation

2.2.2. Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Practical demonstrations

2.2.3. References
• “Best Practices for Forensic Video Analysis (Section 16)” SWGIT
• The Principles and Practice of CCTV 3rd Ed. - Online version

2.2.4. Assessment
• Study questions (oral, written)
• Practical test

2.3 Image Processing

2.3.1. Objectives
• Awareness of colour space conversions
• Knowledge of different image file formats and advantages and disadvantages
• Knowledge of image histograms and RGB colour values
• Knowledge of colour channel separation
• Knowledge of the advantages and limitations of relevant image filters and
adjustments and their impact upon image quality
• Knowledge of image enhancement workflows and the impact of the ordering of
filters on image quality
• Ability to work lossless/uncompressed to maintain image quality
• Ability to apply basic image adjustments and filters such as levels, contrast,
sharpening, de-blurring and noise reduction
• Ability to crop, rescale and rotate images using different interpolation algorithms
• Ability to sufficiently document image processing steps applied to an image

28/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

2.3.2. Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Practical demonstrations

2.3.3. References
• “Best Practices for Forensic Video Analysis (Section 16)” SWGIT
• Cambridge in Colour Tutorials
• ENFSI BPM for Forensic Image and Video Enhancement (ENFSI-BPM-DI-02)
• ASTM E2825 − 12 Standard Guide for Forensic Digital Image Processing

2.3.4. Assessment
• Study questions (oral, written)
• Practical test

29/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

3. IMAGE COMPARISON METHODS AND EVALUATION

3.1 Methods

3.1.1 Objectives
• Knowledge of the Analyse, Compare, Evaluate and Verify (ACE-V) process for
comparison
• Knowledge of the morphological approach and use of feature checklists and
associated limitations
• Knowledge of feature classification and associated limitations
• Knowledge of photo anthropometry and associated limitations
• Knowledge of superimposition and overlays and associated limitations
• Knowledge of the benefits of independent verification and peer-review of observations

• Ability to follow a reliable, repeatable and transparent process for the comparison of
images
• Ability to sufficiently document the process of comparison to ensure transparency
• Ability to assess the suitability of an image or video for image comparison

3.1.2 Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Review of example cases
• Supervised casework
• Clarification on questions
• Discussion

3.1.3 References
• “Best Practices for Forensic Photographic Comparison (Section 16)” SWGIT
• “Facial Comparison Overview (Section 2)” FISWG
• “Guidelines for Facial Image Comparison Methods (Section 6)” FISWG
• R. Moreton, J. Morley, Investigation into the use of photoanthropometry in facial
image comparison. Forensic Science International 212 (2011) 231–237
• K. F. Kleinberg, P.Vanezis, Variation in proportion indices and angles between
selected facial landmarks with rotation in the Frankfort plane, Med. Sci. Law (2007)
Vol. 47, No. 2
• Kleinberg, K.F., Vanezis, P., Burton, A.M., Failure of anthropometry as a facial
identification technique using high-quality photographs, Journal of Forensic Science,
52, 779-783, 2007
• A., A. MCNEILL and D. WHITE, In the Dock: Chimeric Image Composites Reduce
Identification Accuracy, Applied Cognitive Psychology, Appl. Cognit. Psychol. 26:
140–148 (2012)
• S. Ritz-Timme et al. A new atlas for the evaluation of facial features: advantages,
limits, and applicability, Int J Legal Med (2011) 125:301–306
• A. Towler, D. White, R. Kemp, Evaluating training methods for facial image
comparison: The face shape strategy does not work, Perception (2014) 43 214-218
• R. Kemp, , D. White, N. Menon, A. Towler, Detecting identity fraud in passport
applications: Improving human operator performance through better understanding
of unfamiliar face processing (2013), University of New South Wales
• “Facial Image Comparison Feature List for Morphological Analysis” FISWG

30/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

3.1.4 Assessment
• Study questions (oral, written)
• Critical review of case examples
• Proficiency tests
• Test examinations and mock trial

3.2 Contextual and Cognitive Bias

3.2.1 Objectives
• Awareness of the difference between contextual and cognitive bias

• Knowledge of the various sources of contextual and cognitive bias

• Knowledge of the impact of bias on forensic examinations
• Knowledge of strategies to mitigate bias, such as ACE-V and sequential unmasking

• Ability to effectively manage contextual case information to mitigate bias

3.2.2 Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Review of example cases
• Clarification on questions
• Discussion

3.2.3 References
• I. D. Dror, S. A. Cole, The vision in “blind” justice: Expert perception, judgment, and
visual cognition in forensic pattern recognition, Psychonomic Bulletin & Review
2010, 17 (2), 161-167
• I. D. Dror, et al., Contextual information renders experts vulnerable to making
erroneous identifications, Forensic Science International 156 (2006) 74–78
• I. D. Dror et al., Cognitive issues in fingerprint analysis: Inter- and intra-expert
consistency and the effect of a ‘target’ comparison, Forensic Science International
208 (2011) 10–17
• R. Heyer, C Semmler, Forensic confirmation bias: The case of facial image
comparison, Journal of Applied Research in Memory and Cognition 2 (2013) 68–70
(Forensic Debate: Commentary)
• B. Found, J. Ganas, The management of domain irrelevant context information in
forensic handwriting examination casework, Science and Justice 53 (2013) 154–158
• D. E. Krane et al. Sequential Unmasking: A Means of Minimizing Observer Effects in
Forensic DNA Interpretation, J Forensic Sci, July 2008, Vol. 53, No. 4

3.2.4 Assessment
• Study questions (oral, written)
• Critical review of case examples

3.3 Evidence Evaluation

3.3.1. Objectives
• Awareness of distributions (i.e. Gaussian distribution, binomial distribution)
• Awareness of population data and sampling

31/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

• Awareness of Statistics (e.g. probability and calculating mean, median and typical
value)

• Knowledge of the Frequentist and Bayesian approach to evidence evaluation and

their associated limitations
• Knowledge of hypothesis testing
• Knowledge of evidence evaluation and various pitfalls (e.g. prosecutors and the
defenders fallacy, independence assumption and another match error)

• Ability to choose the most suitable approach for evidence evaluation on a case by
case basis
• Ability in presenting the level of conclusion from a comparison and any associated
error-rate
• Ability in establishing appropriate hypotheses and testing them in case work

3.3.2. Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Review of example cases
• Clarification on questions
• Discussion

3.3.3. References
• ENFSI Guideline for Evaluative Reporting in Forensic Science, Strengthening the
Evaluation of Forensic Results across Europe (STEOFRAE), Approved Version 3.0
• Introduction to Statistics for Forensic Scientists, Lucy, D. 2005, John Wiley and Sons
LTD
• Jackson, G. et al. Communicating the Results of Forensic Science Examinations
(November 8, 2015); Penn State Law Research Paper No. 22-2015. Available at
SSRN: http://ssrn.com/abstract=2690899
• Communicating and Interpreting Statistical Evidence in the Administration of
Criminal Justice 1. Fundamentals of Probability and Statistical Evidence in Criminal
Proceedings - Royal Statistical Society
• Communicating and Interpreting Statistical Evidence in the Administration of
Criminal Justice 3. The Logic of Forensic Proof: Inferential Reasoning in Criminal
Evidence and Forensic Science - Royal Statistical Society
• Communicating and Interpreting Statistical Evidence in the Administration of
Criminal Justice 4. Case Assessment and Interpretation of Expert Evidence - Royal
Statistical Society

3.3.4. Assessment
• Study questions (oral, written)
• Critical review of case examples
• Proficiency tests
• Test examinations and mock trial

32/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

4. FACIAL ANATOMY AND FEATURES

4.1 General anatomy

4.1.1. Objectives
• Awareness of the bones of the skull and how they affect facial appearance
• Awareness of the genetic components relating to facial appearance, including
variation between populations and familial similarity (e.g. identical twins)
• Awareness of the muscles of expression
• Awareness of common types of marks and blemishes
• Awareness of the commonality of features within and between relevant populations
• Awareness of the variation of features between males and females (sexual
dimorphism)
• Awareness of acquired features such as scars, piercings and tattoos

• Knowledge of how facial expression affects the appearance of facial features and
causes wrinkles
• Knowledge of the appearance of facial features (e.g. ears, mouth, nose and eyes)
and associated sub-features (e.g. tragus, antihelix and helix of the ear)
• Knowledge of the impact of pose and perspective on facial appearance

• Ability to describe and compare features of the face observed in an image

• Ability to determine the commonality of a face including acquired features

4.1.2. Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Practical demonstrations
• Supervised case work

4.1.3. References
• Forensic Anthropology Training Manual, 2nd Ed. Burns, K. R. 1999. Pearson
Prentice Hall
• Forensic Art and Illustration, Taylor, K. T. 2000. CRC Press
• Claes, P. et al. 2014, Modelling 3D Facial Shape from DNA. PLoS Genetics, 10 (3)
• Human Anatomy and Physiology 6th Ed. Marieb, E. N. 2004. Pearson
• “Anatomy of the Human Body”, Gray, H. 20th Ed.
• Anatomy of the Face and Neck, Prendergast, P. M. in: Cosmetic Surgery, Shiffman,
M. A. and Di Giuseppe, A. (eds), 2012, Springer-Verlag
• Dermatology, Braun-Falco, O. 2000, Springer Science and Business Media
• Nurhudatiana A. et al. 2015, Relatively Permanent Pigmented or Vascular Skin
Marks for Identification: A Pilot Reliability Study, Journal of Forensic Science
• “Facial Image Comparison Feature List for Morphological Analysis” FISWG

4.1.4. Assessment
• Study questions (oral, written)
• Practical test
• Test examinations

33/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

4.2 Ageing and environmental changes

4.2.1. Objectives
• Awareness of the facial changes related to juvenile development
• Awareness of the facial changes related to adolescence
• Awareness of the facial changes related to degenerative ageing
• Awareness of facial changes due to environmental (extrinsic) factors e.g. weight,
lifestyle, diet, drug use
• Awareness of acquired features e.g. scars, tattoos
• Knowledge of the permanence and transience of different facial features over time
• Ability to determine age related and environmental differences between facial
images

4.2.2. Modes of Instruction – Training Aids

• Studying of suggested references
• Classroom based learning
• Practical demonstrations
• Supervised case work

4.2.3. References
• Farkas, J. P. et al. 2013, The Science and Theory behind Facial Aging, PRS GO
• Farkas, J. P. et al. 1992, Growth Patterns of the Face: A Morphometric Study, Cleft
Palate-Craniofacial Journal, 29 (4)
• Albert, A. M. et al. 2007, A review of the literature on the aging adult skull and
face: Implications for forensic science research and applications, Forensic Science
International, 172, 1-9
• The Anatomic Basis of Midfacial Aging, Wulc, A. E. et al. in: Midfacial Rejuvination,
Harston, M. E. (eds), 2013, Springer Business and Media
• Guyuron, B. et al. 2009, Factors Contributing to the Facial Aging of Identical Twins,
Plastic and Reconstructive Surgery, April, 1321-1331
• Rexbye, H. et al. 2006, Influence of environmental factors on facial ageing, Age and
Ageing, 35, 110-115

4.2.4. Assessment
• Study questions (oral, written)
• Practical test
• Test examinations

5. COURTROOM TESTIMONY

5.1 Objectives
• Awareness of the environment of a courtroom and the functions of a criminal
proceeding

• Knowledge of the requirements of an expert witness presenting evidence in court

• Ability to present expert testimony and defend findings

• Ability to present evidence as an impartial expert witness and assist the court

34/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

5.2 Modes of Instruction – Training Aids

• Observation of court testimonies by experienced experts
• Practical exercise (min-mock trials) based on case-studying
• Clarification on questions
• Discussion

5.3 References
• National legislation
• Relevant case law

5.4 Assessment
• Study questions
• Practical exercise in a simulated environment of a courtroom testimony:
• Direct questioning
• Cross-examination

35/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

APPENDIX B: FACTORS AFFECTING FACIAL APPEARANCE

Digital imagery comes from a wide range of sources, such as CCTV, smart devices and social
media. Imagery from uncontrolled sources is frequently used for facial image comparison (FIC),
which may not be of optimal quality. Image quality factors can impact significantly on the reliability
of FIC. This appendix will explain some of these factors and the impact they have on facial
appearance. The assessment of image quality is currently a subjective process and the factors
that affect the quality of one image may not be significant in another image, therefore examiner
should consider multiple factors when determining whether an image is suitable for FIC.

Image factors
Image factors that can alter facial appearance include, but are not limited to:
• Image resolution/distance from camera
• Image compression
• Aspect ratio
• Lighting
• Occlusion
• Camera angle
• Image/lens distortions
• Number of available images
• Date an image was captured

The suitability of an image for FIC may be impacted by one, some or all of the above factors.
The examiner should consider each factor individually and then determine the combined impact
of all the factors on facial appearance. The image quality factors described in this document
are not exhaustive. The examiner may observe artefacts or anomalies in an image that are not
described here.

Note: Facial comparison and facial recognition are two distinct human processes. Often it
is possible to recognise a known face from a low quality image that is unsuitable for facial
comparison. Examiners should be aware of the difference between comparison and recognition
and consider this when assessing the suitability of an image for facial identification (recognition
or comparison).

Image Resolution
The resolution of an image is determined by the number of pixels within that image, the greater
the number of pixels the higher the resolution. Images of a higher resolution will generally
contain more detail than low resolution images, therefore image resolution is the significant
factor affecting facial detail.

If there is insufficient detail to determine the basic shape of facial features that image may be
unsuitable for any method of facial identification.

Enlarging an image beyond its full resolution will not add any more detail to the image, instead
the resulting image may appear blocky or blurry depending upon the method of enlargement
used (e.g. nearest neighbour or bilinear interpolation respectively).

Examiners should consider the resolution of the face in an image rather than only the resolution
of the entire image; i.e. if the subject of interest is a substantial distance from the camera in a
very high resolution image the number of pixels that occupy the face will be low causing a lack

36/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

of facial detail, conversely if the subject of interest is very close to the camera in a low resolution
image facial feature detail may still be discernible.

Figure B1 A low resolution image enlarged to give a blocky result (left) and a blurred result
(centre) compared to the higher resolution original (right)

Questions to consider:
• Can only the basic shape and outline of facial features be determined?
• Does the image appear visibly blocky or blurry due to a lack of pixels?
If the answer is yes to either of the above questions the image may be of limited use for FIC.

Compression
Compression refers to any method used to reduce the data size of a file. Compression may
be lossless, whereby the original file can be recovered and no data is lost (e.g. TIFF files) or
lossy, whereby file sizes can be reduced significantly but may result in a loss of image detail
(e.g. JPEG, MP4). This loss of detail cannot be recovered. Compression can significantly
reduce facial detail even in high resolution images. Highly compressed images will have visible
compression artefacts such as macroblocks or edge artefacts. Compression algorithms may
also subsample colour information leading to chromatic aberrations within an image. Examiners
should be aware of the impact of image compression on facial detail. Resaving imagery in
a compressed format such as a .JPG or .MP4 file will recompress the image and potentially
further reduce facial detail. Where possible images should be resaved in a lossless (e.g. TIFF)
or uncompressed format (e.g. bitmap).

Figure B2 Example of a high resolution image (left) recompressed as a JPEG (right) - note the
visible macroblocks and loss of detail (e.g. creases and marks)

37/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Questions to consider:
• Are compression artefacts visible and contributing to a loss of facial detail?
If the answer is yes to the above question the image may be of limited use for FIC.
• Has the imagery been converted from its original format?
If the answer is yes to the above question the original imagery should be sought as
this may be of a higher quality.

Aspect Ratio
Display aspect ratio refers the proportions of an image as shown on a monitor. It is typically
reported as ratio of the width of the image in proportion to its height (e.g. 4:3, 16:9).
Distortions to the aspect ratio of an image can significantly impact facial proportions. Correct
display aspect ratio can be hard to determine if an image is cropped.

Distortion to video can also occur due to an incorrect pixel aspect ratio. Certain sources of
video, such as standard definition CCTV video, require rectangular, non-square pixels to be
viewed at the correct display aspect ratio. Computer monitors typically display video using
square pixels which will result in the video appearing slightly squashed horizontally. For
example, if standard definition PAL CCTV is displayed using square pixels the display aspect
ratio will be 5:4 rather than the correct 4:3. A correction to the resolution of the imagery is
required to achieve a display aspect ratio of 4:3. Without the correction the proportions and size
of facial features may appear distorted.

Figure B3 Standard definition with a 5:4 display aspect ratio (left) compared to the same image
shown at the correct 4:3 display aspect ratio (right)

Questions to consider:
• Does the aspect ratio of the image appear distorted?
If the answer is yes to the above question the examiner should attempt to correct the aspect
ratio if the native resolution of the image is known.
• Is the imagery a copy of the original and/or cropped?
If the answer is yes to the above question the original imagery should be sought as this
may be at the correct aspect ratio.

Exposure
The lighting within a scene may significantly impact the visibility of facial detail. If an image is
too bright (overexposed) areas of the face may be burnt out and unrecoverable. If an image
is too dark (underexposed) there may be poor contrast limiting facial detail. Adjustments to

38/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

brightness and contrast may recover detail in some circumstances. Image processing, such as
brightness and contrast adjustment should be performed by competent examiners.

Figure B4 Examples of an underexposed image (left), well exposed image (centre) and
overexposed image (right) - note the loss of detail in the dark and light areas

Questions to consider:
• Are parts of the face underexposed and/or overexposed?
If the answer is yes to the above question the image may be of limited use for FIC. Competent
examiners should attempt to recover detail by image processing.

39/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Light Sources
Different light sources emit light in different wavelengths of visible light and may introduce colour
casts into an image. Cameras must have a correctly calibrated white balance to display colour
correctly in captured imagery. If the white balance is not calibrated correctly the skin tone of
a subject may not be accurately represented. Examiners should avoid attempting to correct
the colour of uncontrolled imagery as the results may not be accurate if the original capture
conditions are not known. Examiners should also be cautious in assessing skin colour from
uncontrolled imagery due to variation under different light sources.

Figure B5 Images of the same individual captured under fluorescent lighting with different
camera white balance settings

Near infrared light sources are commonly used in CCTV systems to record video in low light
conditions. Under near infrared light some facial marks, such as freckles, and fine creases may
not be visible and the texture of the skin may appear smoother.

Ref code: ENFSI-BPM-DI-01 Issue No. 001 Page: 48 / 59

40/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

Figure B6 Images of the same individual captured under a visible light source and a near
infrared light source

Figure B7 Images of the right eye of the same individual captured under a visible light source
and a near infrared light source

Occlusion
Facial features may be occluded/obscured from view by other objects within an image, such as
clothing. The higher the number facial features occluded from view the less reliable the image
may be for FIC.

Questions to consider:
• Is there occlusion of facial features in the image?
If a high number of facial features are occluded the image may be of limited use for FIC.

Ref code: ENFSI-BPM-DI-01 Issue No. 001 Page: 49 / 59 41/50

Ref code: ENFSI-BPM-DI-01 Issue No. 001 Page: 49 / 59
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Camera Angle and Pose

Uncontrolled imagery such as CCTV is frequently captured at various angles. In some
imagery (e.g. elevated cameras at an acute angle) facial features may be obscured from view
by the angle of the camera, limiting the suitability of the imagery for FIC. Orientation of the
subject head within the image may also vary, which may also occlude features from view. The
questioned and reference imagery should be at as closely matching pose and camera angle
as possible. For substantial differences in camera angle, such as between a frontal image and
profile image, or pose, facial comparison may not be possible.

Figure B8 The same individual captured at different camera angles

Figure B9 Right ear of the same individual captured at a different pose

Questions to consider:
• Is the camera angle or pose obscuring facial features from view?
If the answer is yes to the above question the image may be of limited use for FIC. If the image
is taken from moving footage the examiner should examine the video to look for additional
imagery.
• Are the images being compared captured at a significantly different camera angle or
pose?
If the answer is yes to the above question the image may be of limited use for FIC. The
examiner should seek additional imagery at a better comparative angle.

42/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

Figure B10 Variation in the appearance of the same individual at different camera angles

43/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Blur
Blur is present in almost all imagery to some degree. If an image is sufficiently blurred facial
feature detail will be distorted or may not be visible, which will negatively impact on a FIC
examination. Image processing methods can be used to sharpen a blurred image by increasing
contrast at edges or by attempting to deconvolve the original blurring process. However
these techniques frequently introduce artefacts into an image that may be misleading during
comparison. Blur can occur due to an improperly focussed camera lens (focal blur) or due to
motion of the subject or a low shutter speed (motion blur).

Figure B11 The same individual captured with increasing focal blur

Figure B12 The same individual captured with increasing motion blur

Questions to consider:
• Is blurring within the image affecting the appearance of facial feature detail?
If the answer is yes to the above question the image may be of limited use for FIC. Competent
examiners should attempt to recover detail by image processing.

44/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

Perspective/Lens Disortion
Images captured when the subject is close to the camera can distort the relative size and
position of facial features. This distortion causes features at the centre of the image to appear
magnified in relation to features at the edge of the images, for example in a facial image the
nose and distance between the eyes may appear larger and the ears may appear smaller.

Figure B13 Images of the same individual captured at different distances from the same camera

Distortion can also be caused by the camera lens. A common example is lens barrelling which
can occur with wide angle lenses. With lens barrelling there may be a pronounced curvature to
features at the edge of the image, this curvature will diminish towards the centre of the image. If
imagery suffers from lens barrelling images used for FIC should be selected when the subject is
in the centre of the image.

Figure B14 Example of lens barrelling distortion in an image

45/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

Questions to consider:
• Is distortion within the image affecting the appearance of facial feature detail?
If the answer is yes to the above question the image may be of limited use for FIC. If the
distortion is caused by the camera lens competent examiners may attempt to recover detail by
image processing.

Expression
The features of the face are dynamic and can be temporarily altered by expression and
contraction of the underlying musculature of the face. Facial expression may distort the size and
shape of facial features, obscure fine feature detail, like marks, from view and cause temporary/
permanent creases at areas of repeated use.

Questions to consider:
• Is there a difference in facial expression between the questioned and reference
imagery?
If the answer is yes to the above question the imagery may be of limited use for FIC. Additional
imagery should be sought with similar facial expression.

Growth and Ageing

As individuals grow and age facial features can change significantly. The growth of juveniles is a
relatively controlled process as the features of face and skull adapt to accommodate functional
changes e.g.:
• Early growth of the cranium to accommodate increases in brain size
• Later growth of the mid face and mandible to accommodate the teeth and growth of
the tongue
During growth facial features can change significantly over a short space of time.

Adult ageing follows a similar general pattern but the rates of ageing vary significantly between
different individuals and can be influenced by environmental factors (e.g. alcohol consumption,
smoking, UV exposure). Adult ageing occurs at a slower rate than juvenile growth. Some
common consequences of aging processes are:
• Hair loss and changes in hair pigmentation
• Sagging of soft tissues
• Increased number and depth of facial lines at areas of repeated muscle use and
facial fat pad junctures
• Decrease in density of fatty tissue under and descent of facial fat pads
• Increased number of hyper pigmented spots

Questions to consider:
• Is there a significant difference in time between the questioned and reference
imagery?
If the answer is yes to the above question the imagery may be of limited use for FIC. Additional
imagery should be sought within a closer age range. The length of time where significant
differences occur will generally be less for juveniles than adults.

Environmental Differences (e.g. weight change, cosmetic surgery, illnesses etc.)

The appearance of facial features can be strongly influenced by environmental factors,
including, but not limited to: weight change, cosmetic surgery, dietary changes, disease, injury,
surgery etc.

46/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

Questions to consider:
 Is there evidence of environmental factors altering the appearance of facial features?
If the answer is yes to the above question the imagery may be of limited use for FIC. Additional
questioned and/or reference imagery should be sought within a closer time frame, or additional
relevant information concerning the subject’s medical history or lifestyle.

47/50
BPM for Facial Image Comparison ENFSI-BPM-DI-01 (vs.01)

APPENDIX C: OVERVIEW OF COMPARISON METHODS

Holistic Face Processing

Holistic comparison is the process by which we naturally identify people. This process is highly
accurate when the faces belong to familiar people (e.g., friends, colleagues, celebrities), but
highly inaccurate when the faces belong to unfamiliar people4. Results from unfamiliar face
matching tasks show large individual differences for holistic face processing– some people
perform remarkably well, whereas others can perform at chance levels5,6. Given that only
unfamiliar faces are encountered in forensics, holistic comparison is only recommended
where other more effective methods are not available (e.g. in quick decision, high throughput
environments such as border crossings), and when the observer has demonstrated suitable
proficiency on unfamiliar face identification tasks. Even when using other facial comparison
methods it is not possible to completely disengage holistic processing.

Photo Anthropometry
Photo anthropometry is a metric approach to facial image comparison. Various soft tissue
landmarks are identified on the face and lines are overlaid onto the images at the determined
location of the landmarks. The distances between the various landmarks are then compared
between the two facial images to determine the level of similarity or dissimilarity. This may be
determined simply by visual inspection or the ratios of the distances between the landmarks
may be compared. Although photo anthropometry is described as a metric technique it is
subjective in nature. The landmarks are often located by eye, which can cause issues when the
imagery is limited in terms of resolution and lighting and can have poor repeatability between
different individuals. Also the threshold for what constitutes a similar or dissimilar measurement
is arbitrary as other factors such as camera angle, resolution, expression and pose may
introduce differences in the facial proportions of the same individual. Empirical studies7,8 have
demonstrated this technique is unreliable with poor repeatability. Photo anthropometry is not
recommended for FIC using uncontrolled imagery.

Superimposition
Superimposition is the combining of two facial images to highlight potential similarities or
dissimilarities. The images may be combined by using a reduced opacity overlay, combining
certain parts of the face within another facial image (chimeric images) or blinking quickly
between two images. The level of agreement/disagreement is determined subjectively by the
examiner. Studies9,10 have demonstrated that the use of chimeric, blinked and overlaid images
4
Bruce, V., Henderson, Z., Newman, C., & Burton, A. M. (2001). Matching identities of familiar and
unfamiliar faces caught on CCTV images. Journal of Experimental Psychology: Applied, 7(3), 207-218.
5
Burton, A. M., Wilson, S., Cowan, M., & Bruce, V. (1999). Face recognition in poor-quality video:
Evidence from security surveillance. Psychological Science, 10(3), 243-248.
6
Bruce, V., Henderson, Z., Greenwood, K., Hancock, P. J. B., Burton, A. M., & Miller, P. (1999).
Verification of face identities from images captured on video. Journal of Experimental Psychology:
Applied, 5(4), 339-360.
7
R. Moreton and J. Morley, “Investigation into the use of photoanthropometry in facial image
comparison,” Forensic Sci. Int., vol. 212, no. 1–3, pp. 231–237, 2011.
8
K. F. Kleinberg and P. Vanezis, “Variation in proportion indices and angles between selected facial
landmarks with rotation in the Frankfort plane.,” Med. Sci. Law, vol. 47, no. 2, pp. 107–16, 2007
9
A. Strathie, A. Mcneill, and D. White, “In the Dock: Chimeric Image Composites Reduce Identification
Accuracy,” Appl. Cogn. Psychol., vol. 26, no. 1, pp. 140–148, 2012.
10
A. Strathie and A. McNeill, “Facial Wipes don’t Wash: Facial Image Comparison by Video
Superimposition Reduces the Accuracy of Face Matching Decisions,” Appl. Cogn. Psychol., vol. 30,
no. 4, pp. 504–513, 2016.

48/50
ENFSI-BPM-DI-01 (vs.01) BPM for Facial Image Comparison

reduced matching accuracy compared to full facial images. The studies also found that chimeric
images created a bias towards false positives and made differences less apparent, indicating
that superimposition can potentially be misleading. Superimposition is not recommended for FIC
using uncontrolled imagery.

Morphological Comparison
Morphological comparison is a subjective process where facial features and/or regions of the
face are observed and compared to determine apparent differences and similarities. Typically
a pre-determined list of features is used in order to structure and document the comparison.
There are two approaches to morphological facial comparison; feature classification and feature
comparison.

Feature Classification
Feature classification involves classifying facial features according to pre-defined categories
(e.g., round face shape, almond eyes, bulbous nose etc.), and basing an identification decision
on the correspondence of these categorisations. However, facial features are not discrete
variables but continuous, so classifying facial features in this way discards an enormous amount
of potentially useful information. Research demonstrates that this technique does not improve
accuracy, is unreliable and can be misleading11,12. Feature classification is not recommended
for FIC. Terminology used in feature classification can be useful for describing observations
in feature comparison (see below) but examiners should not base decisions solely upon an
assigned category or classification.

Feature Comparison
Feature comparison is a form of morphological analysis whereby facial images are compared
feature by feature methodically using a feature checklist and observations are made as to
the similarity/dissimilarity of the individual features. As for feature classification there is no
standardised list of facial features however FISWG have developed a detailed facial feature
checklist document, which if widely adopted by the community will provide some degree of
standardisation in examinations. The features (described as “components”) are each broken
down detailed sub-component characteristics. Empirical research has shown that conducting
a systematic feature-by-feature comparison improves identification accuracy when the facial
images are a true match, compared to results from participants who reached a decision without
a feature list (see above for holistic comparison)13. The study also found that for forensic
examiners, the ears, nose, and scars and blemishes were most diagnostic of whether the
faces showed the same person or different people. The results also confirm reports by facial
comparison examiners that ears are often the most useful feature for comparison. Given the
evidence in support of this comparison method, feature comparison is recommended for FIC. It
is also recommended that the FISWG facial feature checklist be used in such analyses.

For all comparison methods, including morphological analysis, laboratories should conduct their
own validation studies of their method using ground truth material that represents the type(s) of
imagery commonly encountered in casework. Laboratory staff should also undertake scheduled
proficiency testing or inter-laboratory testing to evaluate their performance.
11
A. Towler, D. White, and R. I. Kemp, “Evaluating training methods for facial image comparison: The
face shape strategy does not work,” Perception, vol. 43, no. 2–3, pp. 214–218, 2014.
12
S. Ritz-Timme et al., “A new atlas for the evaluation of facial features: Advantages, limits, and
applicability,” Int. J. Legal Med., vol. 125, no. 2, pp. 301–306, 2011.
13
A. Towler, D. White, R. I. Kemp, A. Towler, D. White, and R. I. Kemp, “Evaluating the Feature
Comparison Strategy for Forensic Face Identification,” J. Exp. Psychol. Appl., 23 (1), 2017.

49/50
Best Practice Manual for Facial Image Comparison
ENFSI-BPM-DI-01
Version 01 - January 2018