Integration of the Optimal Gabor Filter Design and Local Binary Patterns for Texture Segmentation

This paper presents a novel approach for multi-textured image segmentation based on an effective texture feature formation. Partial texture features are derived from an optimal Gabor filter selected by Immune genetic algorithm (IGA), which aims at maximizing the discrimination power among multi-textured regions in images. Then the proposed method integrates the texture features related to the Gabor filter with a local binary pattern, an effective texture descriptor at low computational cost, to overcome weakness of a single frequency component of the filter. Finally, a K-nearest-neighbor classifier is used for the tasks of multi-textured image segmentation. The integration of optimal Gabor filtering and local binary pattern provides a new solution for the tasks of multi-textured images. The experimental results demonstrate the effectiveness of the proposed approach.