Entropy Based Local Binary Pattern Operator: A Texture Segmentation Approach

Abstract

Texture segmentation is a critical task in image analysis with numerous applications in computer vision. This paper proposes an efficient approach for unsupervised texture segmentation that leverages features extracted from the Entropy Based Local Binary Pattern (LBP) Operator, combined with Fuzzy-c-Means (FCM) and K-Means clustering methods enhanced by spatial information. The proposed method reconstructs the texture mosaic image using LBP to capture detailed local texture information. Entropy values, which quantify the randomness and texture details in the LBP image, are then extracted. These entropy values are clustered using FCM and K-Means algorithms, which are modified to incorporate spatial context for improved segmentation accuracy. The effectiveness of the approach is evaluated across various texture databases, demonstrating superior performance compared to existing methods. The proposed algorithm excels in capturing and segmenting texture information, offering a more comprehensive representation of texture characteristics. Experimental results show that this proposed method achieves higher efficiency and accuracy, making it a robust choice for texture image segmentation.