Study Of Texture Image Segmentation Based On The Quotient Space

Artificial intelligence researchers have been studying human ability of solving problem for a long time, and lots of models were presented. Among them, the quotient space model is classic. Based on the human thinking and granularity concept, the problem is analyzed from different points of view and at different levels. According to granularity synthesis principle, the different granularities are synthesized, then the solution is found.The image segmentation is an important prerequisite technique in image analysis and computer vision system, which is a basic and difficult problem in the studying field of computer vision. In the past few decades, numerous researches have been carried out and many different image segmentation algorithms were exploited. However there was no good solution for the image segmentation problem yet. It is a challenging problem to grope for the image processing methods based on artificial intelligence and gear to human visual character.Before the quotient space granularity concept is proposed, the image-processing problem based on the quotient space has been studied. Moreover, it is mainly explored that the quotient space theory is applied to the field of texture image segmentation. This paper is intended to make a union thinking method and algorithm frame in image analysis and processing. The main contributions of the paper are as follows.1. After the granularity concept, rough set, fuzzy set and other granularities computing theory are studied and compared, the image segmentation concept and the segmentation methods are analyzed completely. Through the work above, the quotient space granularity theory in image segmentation is presented, which is used to generalize the image segmentation algorithms, and then the image segmentation theory frame based on granularity computing is built.2. The texture image segmentation algorithm based on the DFT pixel 8-neighboroperator is proposed. Firstly, the histograms based on DFT operator of pixel 8-neighbor are formed according to the quotient space model. Secondly, the equivalence relations through the difference between histograms are also formed, which are used to classify pixels. Finally, texture images can be segmented by the clustering method. The experimental results show that the algorithm proposed is valid.3. The texture image' segmentation algorithm based on granularity synthesis technique is put forward. Making use of the human vision perception character and the granularity synthesis theory, the characteristics of directionality and roughness in texture image are extracted, and the different granularities are formed. According to granularity synthesis model, the algorithm of texture image segmentation is successfully implemented, and the experimental results are obvious.4. The texture image segmentation algorithm based on granularity and wavelet transform is proposed. After analyzing and summing up the wavelet theory, especially the second-generation wavelet theory, the conclusion is educed that wavelet and quotient space theory are accordant in essence according to the quotient space approach theory. After the wavelet transform with granularity hierarchical and the synthesis technique being combined, a novel segmentation algorithm in texture image is presented, and the algorithm is verified by experimental results.5. The texture image segmentation algorithm based on the fuzzy granularity theory is brought forward. The quotient space granularity theory is extended into fuzzy quotient space, and then some important conclusions based on the fuzzy quotient space are presented. On the other hand, through combining wavelet with fuzzy soft cluster technique and granularity thinking, the texture image segmentation algorithm based on fuzzy quotient space is put forward, which is valid to confirm multi-texture neighborhood boundary.In summary, the theory and algorithms of the texture image segmentation based on granularity hierarchical and syntheses technique are put forward. Thetexture image segmentation methods based on the wavelet technique and the granularity theory and the methods based on the fuzzy granularity theory are studied. This paper illuminates the significance of quotient space theory and its value in theory and practice, which accelerate the popularization of the quotient space theory and its application in other fields.