Extraction Of Feature For Efficient Analysis And Classification Of Biomedical Image

Abstract

The purpose of this paper is to demonstrate effectiveness and efficiency of different fusion techniques in the analysis and characterization of diseases pattern in medical images. Local features have been combined with global features of images to generate new descriptors or features for efficient characterization and classification of biomedical images. Some of the important research questions that would be specifically addressed in this paper include the following: What are the techniques for feature extractions?

How can the fractal dimension be applied to detect different patterns in medical images? Can the combined features from both local and global features provide robust descriptors of shape/textures for the analysis and classification of images? Answers to these research questions would be illustrated with several experiments. Several methods of feature fusion techniques in the analysis of medical images have been suggested and evaluated in different ways. This paper proposes to develop new descriptors using fusion techniques to address some of the limitations of the existing methods in terms of efficiency, speed of image analysis and error corrections. Results analysis and some important implementation aspects of the new descriptors that could be used to improve the overall classification accuracy have been discussed in this paper. The performance measure of different models has been investigated using combined features for further improvement and analysis in the detection and classification of Emphysema patterns.