Texture Segmentation by a New Variant of Local Binary Pattern

This paper highlights the local binary pattern (LBP) method in the unsupervised texture segmentation task. It has been made into a really dominant measure of image texture, showing outstanding results in terms of computational complexity and accuracy. The LBP operator is a theoretically simple yet very efficient approach for texture analysis. The LBP concept is slightly modified, i.e., instead of considering the center pixel value for generation of binary values, the present paper utilized average of all the eight neighboring pixels of the center pixel. The binary code generated is separated into "Diamond-LBP Code (DLBPC)" and "Corner LBP code (CLBPC)." The proposed new variant local binary pattern (NVLBP) segmentation approach is simple, rotationally invariant and easy to understand. This method also resulted in good segmentation which is noticed from the entropy, standard deviation, contrast, and discrepancy values.