A New Technique for Texture Classification Using Markov Random Fields

This paper proposes, applies and evaluates a new technique for texture classification in digital images. The work describes, as far as possible in a quantitative way, the concept of texture in digital images. Furthermore, we developed an innovative model that allows classifying and characterizing texture in digital images, to be used as a useful tool in noninvasive inspection of visual surfaces. The proposed methodology extracts the statistical order from an image of texture. The extraction of the high statistical order has been made using as a tool Markov Random Fields. The Backpropagation neural net is used for designing a classification module that will serve to test the performance of the configuration histograms, which are based on the statistical order. Furthermore, the research suggests the evaluation of the proposed technique from a qualitative perspective.