Scribd
Upload
49 views2 pages

6 Texture in Biomedical Images: Summary

This document discusses texture analysis methods for biomedical images. It provides an overview of various texture analysis techniques including Markov random fields, co-occurrence matrices, Gabor filters, wavelets, and local symmetry/orientation methods. It also reviews some medical image processing applications that have used texture analysis. Texture is defined as variation within an object at a scale smaller than the overall object. Texture analysis aims to objectively characterize textures regardless of human perception. The challenges include classifying texture patches and simultaneously isolating and characterizing multiple textures within an image.

Uploaded by

chandro pardede
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
49 views2 pages

6 Texture in Biomedical Images: Summary

This document discusses texture analysis methods for biomedical images. It provides an overview of various texture analysis techniques including Markov random fields, co-occurrence matrices, Gabor filters, wavelets, and local symmetry/orientation methods. It also reviews some medical image processing applications that have used texture analysis. Texture is defined as variation within an object at a scale smaller than the overall object. Texture analysis aims to objectively characterize textures regardless of human perception. The challenges include classifying texture patches and simultaneously isolating and characterizing multiple textures within an image.

Uploaded by

chandro pardede
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

6

Texture in Biomedical Images

Maria Petrou

Summary. An overview of texture analysis methods is given and the merits of


each method for biomedical applications are discussed. Methods discussed include
Markov random fields, Gibbs distributions, co-occurrence matrices, Gabor functions
and wavelets, Karhunen–Loève basis images, and local symmetry and orientation
from the monogenic signal. Some example applications of texture to medical image
processing are reviewed.

6.1 Introduction

Texture is variation of the data at scales smaller than the scales of interest.
For example, if we are interested in identifying the human brain in a Mag-
netic Resonance Imaging (MRI) image, any variation in the gray values of the
imaged brain may be thought of as texture. According to this definition, even
variation due to noise may be thought of as texture. In the computer graphics
community this is acceptable, but among image processing and pattern recog-
nition researchers, texture is a property intrinsic to the imaged object and not
something caused by the instrument with which the image has been captured,
like noise is. Once we accept texture to be an intrinsic property of the imaged
object, then texture becomes a valuable cue in relation to the recognition of
this object. For example, the texture created by the sulci helps us to identify
the brain as such.
The most important characteristic of texture is that it is scale dependent.
Different types of texture are visible at different scales. For example, if we
look at a section of the human brain through a microscope, we are going to
see different structure of the tissue than the sulci mentioned above. In order
to be able to use texture to identify different types of tissue or different human
organs, we must be able to measure it in a repeatable and reliable way. In
other words, it is necessary to be able to characterize texture in an objective
way, independent of human perception and visual abilities. In this chapter,
we first present a brief overview of texture quantification methods and discuss

T.M. Deserno (ed.), Biomedical Image Processing, Biological and Medical Physics,
Biomedical Engineering, DOI: 10.1007/978-3-642-15816-2 6,

c Springer-Verlag Berlin Heidelberg 2011
158 M. Petrou

the peculiarities of texture analysis in relation to biomedical images. Then


some examples of successful use of texture cues in medical applications are
presented.
The simplest problem one tries to solve using texture is to identify the class
of a texture swatch. In this case, one assumes the availability of a library of
textures of different classes, all captured under the same imaging conditions as
the texture swatch in question. The issue then is to quantify the query texture
in the same way as the library textures have been quantified, and compare
the extracted values with those of the library textures, to assess similarity or
divergence. The query texture is identified as of the same class as the texture
in the library with the most similar numerical values.
The second, and most difficult problem one has to deal with, is that of
simultaneous isolation and characterization of the texture. This problem arises
very often, as in the same image, different depicted objects may be character-
ized by different textures. The problem becomes complicated as soon as one
realizes that the borders between the different textures are often diffuse, with
the textures blending gradually into each other, and that the shapes of the dif-
ferent texture patches are not necessarily regular and sometimes the patches
are not even extensive enough for the texture to be easily characterizable by
automatic methods.
Next, we shall discuss these two problems in turn.

6.2 Characterizing the Texture of Swatches

6.2.1 From Grammars to Markov Random Fields

Texture may be characterized by structural or statistical methods. A struc-


tural method identifies a texture primitive pattern and the rules by which this
is repeated to create the viewed appearance. This approach leads to the so
called grammars for texture description. Such methods are of little interest to
biomedical applications, but they tend to be appropriate for man-made tex-
tures that present some regularity. Statistical approaches are of more interest.
Figure 6.1 shows schematically how natural the transition from deterministic
methods to statistical methods is. In this example, we may easily infer in the
first case that a possible texture primitive is a 1 surrounded by four 0s in its
four immediate neighboring positions. In the second case, we may work out
the pattern by inspecting the instantiations of this type of neighborhood we

Fig. 6.1. Deterministic versus probabilistic 1 0 1 0 1 0 1 1 0 1 0 1 0 1


context dependence. Left: It is obvious that the 0 1 0 1 0 1 0 0 1 0 1 0 1 0
1 0 1 0 0 1 2 0 1 0 0 1
missing value should be 1. Right: The missing 0 1 0 1 0 1 0 0 2 0 2 0 1 0
value has probability 58% to be 1 and 1 0 1 0 1 0 1 2 0 1 0 2 0 1
probability 42% to be 2 0 1 0 1 0 1 0 0 1 0 2 0 2 0
1 0 1 0 1 0 1 2 0 1 0 1 0 1

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy