The Classification Study Of Texture Image Based On The Rough Set

Rough Set theory deals with the classification analysis of data tables. The main goal of the rough set analysis is synthesize approximation of concepts from the acquired data. It has been successfully applied in bioinformatics, economics and finance, medicine, web and text mining. In this paper, classification of texture image theory is introduced through the research of rough set theory. The classification of texture image procedures consist of two main steps:feature extraction and feature classification. Both of two steps discussed in this paper as well. The details are shown as follows:(1)In the course of feature extraction, two methods are called gabor filters and gray level co-occurrence matrix (GLCM) are respectively proposed to extract the features of texture image. Gabor filter has combined with optimal resolution of the time domain and frequency domain, in addition, it simulates the visual characteristics of the human visual system well, which capture low and medium frequency textural information precisely. So this paper designs a set of four scales and six directions of the gabor filter to carry on filter processing to the texture image for 24 times, then extracts mean and variance from each output of 48 vectors values as its features. Gray level co-occurrence matrix can capture high frequency textural information better, texture of extraction with good judgment, and the idea is simple, easy to implement, it has strong vitality. This paper extracts 32 features from four directions of GLCM, each direction has eight features which can describe different aspects of texture, such as angular second moment, entropy, contrast and so on. (2) In the course of classification stage, starting with rough set theory as a classifier, then in accordance with the general steps of rough set theory to discriminate, reduce attribute and extract rules. The different groups with discrimination algorithm and attribute reduction algorithm have a great impact on the final classification results. Therefore, use different combination of discrimination algorithm and attribute reduction algorithm to compare with the final result of classification. Finally, comparing the classification results by the number of correctly identified as the evaluation criteria. Found that combination of Boolean reasoning discrimination and mutual information-based algorithm for reduction of knowledge can get a better classification result.Because of the complementarily of rough set theory and support vector machine, based on the view that combining their advantage by some way maybe obtain better results of classification than a single classifier in the overall. This paper uses a multi-classifiers fusion method which combines rough set and support vector machine classifier. The step of the multi-classifiers fusion method is shown as follows:First of all, using the rough set to reduce the dimension of the attribute by attribute reduction algorithm which can remove the redundant information. Secondly, pass the new property to SVM classifier to classify. The multi-classifiers fusion method not only makes full use of the advantage which has excellent ability to reduce attributes of rough set and has a high recognition rate of SVM, but also simplifies the training and classification of SVM classifier, and improve the classification accuracy as well.