Texture Image Segmentation Based On Non-local Methods

Texture image segmentation, as an important research content in the image processing and pattern recognition area, has been a hot spot of research. Because the texture image is different from the simple image, it has much complex texture information; therefore, it is relatively difficult for texture image segmentation. To solve the problem of texture image segmentation, this paper will promote nonlocal operator to the active contour model. It uses the gradient factor of active contour model combined with the nonlocal operator for texture image segmentation; at the same time, it also use R, G, B color space of color image to achieve color texture image segmentation. The main work and innovations are as follows: first, it introduces the basic theory for texture image segmentation, mainly introducing the texture image characteristics and variational image segmentation method.Second, it introduces the image segmentation model based on variational theory, mainly introducing Mumford Shah model and Chan-Vese model of variational image segmentation; at the same time, it also introduces active contour model based on texture decomposition. Third, the traditional active contour model is introduced into the nonlocal gradient of the similarity degree based on image region similarity, texture image segmentation can be implemented. Because the nonlocal operator will result in a long run time, so the grayscale texture image segmentation model which only adds the nonlocal gradient term is modified, the new original gradient is added. At this point it achieves grayscale texture image segmentation and shorten the operation time at the same time.Fourth, because of the color image can be regarded as R, G, B channel image, it extends the modified grayscale texture image segmentation model to color texture images to achieve color texture image segmentation. Compared with the traditional image segmentation method based on image pixel of image processing, the proposed method is based on the image processing of similarity of image area, and this method can effectively identify the region of interest in the texture image. The model not only contains a nonlocal gradient, but also add the new gradient to accelerate the speed. To improve the operation efficiency of the model, this paper uses the Split-Bregman algorithm for it.Finally the effectiveness of the proposed model and algorithm is verified by the experiments.