Feature Learning of Image for Unsupervised Texture Segmentation

Segmentation in specific image class, texture feature extraction plays a vital role. But is time consuming and difficult, to develop novel technique to select features manually. Adapting features automatically for particular class of images, will be helpful and effective. Thus, proposed work contributes in texture segmentation, in absence of large learning datasets for learning features. Cost function is used to develop, learning process for matching segmentation model (Mumford Shah model). The feature learning provides essential piecewise image with constraint feature set of compact jump. It is based on convolution feature learning and segmentation performance respectively. It has been reflected, it is possible to learn features of image patch. The approach is effective and produces better texture segmentation for natural images.