Texture classification using sparse frame based representations

Abstract

In this paper a new method for texture classification, denoted Frame Texture Classification Method (FTCM), is presented. The main idea is that a frame trained to make a sparse representation of a certain class of signals is a model for this signal class. The signal class is given by many representative image blocks of the class. Frames are trained for several textures, one frame for each texture class. A pixel of an image is classified by processing a block around the pixel, the block size is the same as the one used in the training set. Many sparse representations of this test block are found, using each of the frames trained for the texture classes under consideration. Since the frames were trained to minimize the representation error, the tested pixel is assumed to belong to the texture for which the corresponding frame has the smallest representation error. The FTCM is applied to nine test images, yielding excellent overall performance, for many test images the number of wrongly classified pixels is more than halved, in comparison to state of the art texture classification methods presented in [1].