BIOMETRIC IMAGE

RESTORATION

1
IMAGE NOISE

Noise in digital images can arise during image

acquisition (digitization) and transmission
• Imaging sensors can be affected by ambient
conditions
• Environmental interference can be added to an
image during transmission

Image noise is random variation of brightness or

color information in images, and is usually an aspect
of electronic noise

It can be produced by the sensor and circuitry of a

scanner or digital camera

2
IMAGE RESTORATION

Image restoration attempts to restore images that

have been degraded due to digitization and/or various
types of noise

Aim : To identify the degradation process (type of

noise) and attempt to reverse it

Image restoration can be done in :

• Spatial domain, or
• Frequency domain

3
MODELING OF NOISE
We can consider a noisy image to be modelled as
follows:
g(x, y) = f(x, y) * h(x, y) + η(x, y)
where f is the original image, η is the noise
function, h is the degradation function and g is the
resulting noisy image

Image degradation occurs due to loss of information

while storing the acquired image (for example,
conversion of 8-bit image to 5-bit image)

There are many different models for the image

noise term η(x, y):

4
MODELING OF NOISE

There are many different models for the image

noise term η(x, y):

• Gaussian noise
• Rayleigh noise
• Exponential noise
• Gamma noise
• Impulse noise (salt and pepper noise)
• Uniform noise (quantization noise)

5
 NOISE MODELS
Type of Noise Possible Cause
Gaussian noise Electronic circuit noise, sensor noise
due to poor illumination and/or high
temperature (during image
acquisition)
Rayleigh noise Loss of light signal during range
imaging (images captured with depth
camera which uses spread-out beam
of light to record distance of objects)

Exponential noise Loss of light signal during laser

Gamma noise imaging (images captured with
LiDAR which uses point-wise
concentrated light to record distance
of objects)
6
 NOISE MODELS

Type of Noise Possible Cause

Impulse noise Quick transients, such as faulty
(Salt and Pepper noise) CCD/CMOS sensor cells (during
image acquisition)
Uniform noise Caused by digitization or
(Quantization noise) compression of images

7
GAUSSIAN NOISE MODEL

Gaussian Noise :

• Mean :
• Variance :

8
RAYLEIGH NOISE MODEL

Rayleigh Noise :

• Mean :
• Variance : =

9
GAMMA NOISE MODEL

Erlang (Gamma) Noise :

• Mean :
• Variance :

10
EXPONENTIAL NOISE MODEL

Exponential Noise :

• Mean :
• Variance : =

11
UNIFORM NOISE MODEL

Uniform Noise :

• Mean :
• Variance :

12
IMPULSE NOISE MODEL

Impulse (Salt and Pepper) Noise :

• Salt and Pepper noise is added to an image by

addition of both random bright (with probability )
and random dark (with probability ) all over the
image

13
VISUALIZATION OF EFFECT OF NOISE
The test pattern below is ideal for demonstrating the
addition of noise

The following slides will show the result of adding

noise based on various models to this image

Histogram to go here

Original image Histogram

14
VISUALIZATION OF EFFECT OF
NOISE

15
VISUALIZATION OF EFFECT OF
NOISE

16
SPATIAL FILTERING

An image degraded only by additive noise can be

expressed as :
g(x, y) = f(x, y) + η(x, y)
The noise terms generally are unknown, so subtracting
them from g(x, y) to obtain f(x, y) is not an option

Spatial (time domain) filtering is the most commonly

used method for image denoising

Mean and median filters are the most commonly used

spatial filters

17
ARITHMETIC MEAN FILTER

The arithmetic mean filter computes the average

value of the corrupted image, g(x, y), in the area
defined by .

where r and c are the row and column coordinates

of the pixels contained in the neighborhood .

A mean filter smooths local variations in an image,

and noise is reduced as a result of blurring.

18
GEOMETRIC MEAN FILTER
An image restored using a geometric mean filter is
given by the expression :

Here, each restored pixel is given by the product of

all the pixels in the sub-image area, raised to the
power .

A geometric mean filter achieves smoothing

comparable to an arithmetic mean filter, but it tends to
lose less image detail in the process

19
HARMONIC MEAN FILTER
The harmonic mean filtering operation is given by
the expression :

The harmonic mean filter works well for salt noise,

but fails for pepper noise. It does well also with other
types of noise like Gaussian noise.

20
MEDIAN FILTER

The median filter replaces the value of a pixel by the

median of the intensity levels in a predefined
neighborhood of that pixel :

Median filters provide excellent noise-reduction

capabilities, with considerably less blurring than
linear smoothing filters of similar size.

21
GABOR FILTER
The Gabor Filter is a linear filter used in image
processing applications for edge detection, texture
analysis, feature extraction, etc.

These filters have been shown to possess optimal

localization properties in both spatial and frequency
domain and thus are well suited for texture
segmentation problems.

Gabor filters are special classes of band pass filters,

i.e., they allow a certain ‘band’ of frequencies and
reject the others.

22
DEFINITION OF 1D GABOR FILTER
The one-dimensional Gabor filter is defined as the
multiplication of a cosine/sine (even/odd) wave with
a Gaussian window, as follows :

where defines the center frequency, and the spread

(standard deviation) of the Gaussian window.

23
VISUALIZATION OF 1D GABOR FILTER

24
THE COMPLEX GABOR FILTER
 The 1D complex Gabor filter is given by :

Similarly, the 2D complex Gabor filter is given by :

25
VISUALIZATION OF 2D GAUSSIAN

26
VISUALIZATION OF 2D COSINE

27
VISUALIZATION OF 2D GABOR FILTER

28
GABOR FILTER BANK
From the previous slide, we can notice that the
sinusoid has been spatially localized.

In practice to analyze texture or obtain feature from

image, a bank of Gabor filter with number of
different orientation are used.

Consider a fingerprint image which has a pattern of

ridges at various orientations.

Now to highlight or extract out all those patterns we

are going to use a bank of 16 Gabor filters at an
orientation of 11.250

29
FILTER BANK VISUALIZATION

30
RESULT OF GABOR FILTERING

The Gabor filter detects and aims to preserve the

curves in an image, while eliminating random noisy
pixels in between such curves.

For a fingerprint image, the Gabor filter preserves

the curved ridge lines while eliminating any noisy
pixels in the valley region

31
VISUALIZATION OF GABOR FILTERING

(a) Original fingerprint

image

(b) After applying

median filter

(c) After normalization

and binarization

(d) After applying

Gabor filter

32
 REFERENCES
[1] Rafael C. Gonzalez, Richard E. Woods – Digital Image Processing (2017)

[2] Through The Eyes of Gabor Filter :

https://medium.com/@anuj_shah/through-the-eyes-of-gabor-filter
-17d1fdb3ac97

[3] Gabor filters :

http://www.cse.yorku.ca/~kosta/CompVis_Notes/gabor_filters.pdf

[4] Ali Abdul Azeez Mohammad Baker, “Gabor Filter”, International Journal of
Computer and Communication System Engineering (IJCCSE), vol. 1, pp. 92-96, 2014

[5] Understanding Gabor Filter :

https://kgeorge.github.io/2016/02/04/understanding-gabor-filter

[6] Hong, Lin and Wan, Yifei and Jain, Anil, “Fingerprint image enhancement:
algorithm and performance evaluation”, IEEE transactions on pattern analysis and
machine intelligence, vol. 20, pp. 777-789, 1998

33