ABSTRACT

This paper presents a color texture descriptor, Modified Local Binary Pattern (MLBP) based on combined intensity and color information. It establishes the existence of correlation between the intensity and color channel information along with their corresponding neighbors. Bin values of MLBP image histogram are used as color texture descriptors. The classification performance of the proposed color texture descriptor is evaluated on VisTex and Outex datasets, for different color spaces, namely RGB, HSV, YCbCr , La*b* using k-nearest neighborhood classifier. The classification results obtained by the proposed texture descriptor are compared with the results obtained by the conventional Local Binary Patterns (LBP) method. The experimental results indicate that the proposed color texture descriptors enhance the classification performance.