Texture Image Segmentation Based On Region Feature

With the rapid development of information media,image segmentation has a wide range of applications.Texture image segmentation is one of the important research fields.However,the problem of texture image segmentation is still not well resolved because of the variety of image texture features.Therefore,this research has important theoretical and practical significance.In this paper,several effective segmentation models from the perspective of structure tensors and cumulative histograms are proposed,the details are described as follows:?1?The nonlinear extended structure tensor is constructed based on the structure tensor.A scalar texture descriptor T_a is constructed according to the relationship between matrix eigenvalues and each channel of the tensor matrix.This texture descriptor contains texture information and gray information.At the same time,it also considers the correlation among the channels of the matrix.A hybrid texture segmentation model A is constructed by embedding T_a into the traditional GAR model and the gray information I into the traditional Chan-Vese?CV?model.The fast dual method is used to solve the energy minimization problem of model A.?2?Because the distance between histograms can be used to describe the difference of two regions,a texture segmentation model B is proposed based on the d_L? distance between cumulative histograms.A localized version?model C?of segmentation model B is proposed based on the idea of LBF model.The fast dual method is used to solve the model B and model C.Model B and model C can also be solved by Split-Bregman fast iterative technique.Moreover,an adaptive image interpolation algorithm is proposed,which is based on the contrast and the direction angle of the image.The experimental results show that the algorithm overcomes image aliasing.