Texture Classification Using Texton Co-Occurrence Matrix Derived From Texture Orientation

The present paper derived a new co-occurrencematrix based on textons and texture orientation for rotationinvariant texture classification of 2D images. The newco-occurrence matrix is called as Texton and Texture OrientationCo-occurrence Matrix (T&TO-CM). The Co-occurrence Matrix(CM) characterizes the relationship between the values ofneighboring pixels, while the histogram based techniques havehigh indexing performance. If the CM is used to represent imagefeatures directly, then the dimension will be high and theperformance is decreased. On the other hand, if histogram is usedto represent image features, the spatial information will be lost.Texture Classification based on T&TO-CM, integrates color,texture and edge features of an image. The proposed T&TO-CM isused to describe the spatial correlation of textons and textureorientation for texture classification. T&TO-CM can capture thespatial distribution of edges, and it is an efficient texturedescriptor for images with heavy textural presence. The proposedmethod is computationally attractive as it computes differentfeatures with limited number of selected pixels. The experimentalresults indicate the efficacy of the present method over the variousother methods.