文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Optimal Feature Functions on Co-occurrence Matrix and Applications in Tumour Recognition
侵权投诉

Optimal Feature Functions on Co-occurrence Matrix and Applications in Tumour Recognition

The level of quality in segmentation and classification tasks, operating on biomedical digital images, is strongly affected by the ability of the feature functions to map the input structures on “highly separated ” data sets. This level of separation is

Optimal Feature Functions on Co-occurrence Matrix

and Applications in Tumour Recognition

Angelo Zizzari*, Bernd Michaelis* and Günther Gademann**

*Institute for Electronics, Signal Processing and Communications, **Clinic for Radiotherapy

“Otto-von-Guericke” University – P.O. Box 4120, 39016 Magdeburg, Germany

e-mail: zizzari@iesk.et.uni-magdeburg.de

Abstract

The level of quality in segmentation and classification tasks, operating on biomedical digital images, is strongly affected by the ability of the feature functions to map the input structures on “highly separated” data sets. This level of separation is strictly dependent by the “discri-minatory power” of the chosen feature functions, meaning with this expression the property of the functions to assign some very different values to different types of image blocks.

This paper describes a possible theoretical approach addressed to the improvement of this “discriminatory power” of textural features in the particular case of features extracted using a co-occurrence matrix method. After the necessary references to the basic concepts, a formal definition of the “Discrimination Enhancement Problem” is stated. A new theorem is presented, solving the problem of discrimination among three different classes. The chosen application is the problem of tumour recognition in radiographic images of the brain and the experimental results support the effectiveness of this approach.

Keywords : Tumour Detection, Co-occurrence Matrix, Textural Features, Pattern Analysis and Recognition.

1. Introduction

In the scientific literature, texture-based applications, providing segmentation and classification tasks on digital images, are widely reported. The basic assumption of such applications is that different local textures can represent different physical characteristics of different objects. Some examples are described in the field of biomedical informatics [1], where texture measurements give infor-mation about the presence of masses in mammograms; in oceanography [2], where objects, located on the medi-terranean seabed, are accurately detected and classified; in computer vision [3], where specific elements of natural landscapes are extracted and recognized from images.

The way to obtain these textural measures could follow different approaches. The application of a set of con-volution kernels leads to the Laws’ texture energy measures [4]; the extraction of features respect to the fractal dimension is related to the fractal-based texture approach [5]; furthermore, the gradient and edge opera-tors can be employed [6], as well as the autocorrelation functions, local histograms, unsharp masking and median filters [7].

In our previous study [8], a computer system has been designed and developed, in order to detect the presence of a tumour in digital images of the brain and accurately define its borderlines. It was chosen a gray level co-occurrence matrix approach, following the formulation introduced by Haralick et al. [9], which is a well-known statistical method for extracting second order texture information from images. The result was strongly affected by the ability of the feature functions to map the underlying structures on “highly separated” data sets, and this level of separation was strictly dependent by the discriminatory power of the chosen feature functions, meaning with this expression the property of the functions to assign some very different values to different types of image blocks.

Previous significant research, dealing with the objective of improving such a discriminatory power, is described in [10], where a discriminatory matrix is built by measuring the class separability of elements of co-occurrence matrices. This discriminatory matrix is used as a weight-ing matrix, able to increase the contribution of some elements and to decrease the contribution of others, in order to improve the discriminatory power of the classical textural features. In [11], an overview of further deve-lopments is presented, where several adaptive methods are used for the creation of highly discriminatory features.

Our paper describes a possible theoretical approach towards the improvement of such a discriminatory power of textural features. The main objective is the definition of optimal parameters which solve the problem of maximizing this level of separation among classes of different objects. A formal definition of the “Discri-

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email