Texture Classification Based On Support Vector Machine

Texture can be seen everywhere in daily life, at the same time, texture recognition plays an important role in the areas of image processing. The main content of texture recognition includes two aspects: texture features extraction and establish the appropriate classifiers. The main approach of texture features extraction includes four parts, also a lot of relatively mature methods are included; On the other hand, lots of classifiers are used.The main contributions and contents of this thesis are given below:(1)Firstly, Support vector machine is elaborated detailed, its main idea and method are also included. Then Support vector machine based on least squares method is recommended, also it is served as classifiers in this thesis.(2)Secondly, as invariant feature recognition is frequently discussed, so the chief methods about them are summaried in this thesis. Then the affine invariance about texture recognition is proposed with the aid of these methods. The affine transform is converted to translation transform, then DT_CWT is used to extract texture features.(3)Thirdly, as wavelet analysis is an outstanding tool of image processing, so in the last chapter of the thesis, it is elaborated simply. Then Multiscale Geometric Analysis are introduced, also their advantages and disadvantages are demonstrated, then using SWBCT to extract texture features. At last, SVM is used for testing and training with validity capability.