Local line directional neighborhood pattern for texture classification

Abstract

Local binary pattern (LBP) and combination of LBPs have shown to be a powerful and effective descriptor for texture analysis. In this paper, a novel approach to pattern recognition problem namely local line directional neighborhood pattern (LLDNP) for texture classification is proposed. The proposed LLDNP extracts the directional edge information of an image at 0°, 15°, 30°, 45°, 60°, 75°, 90°, 105°, 120°, 135°, 150°, and 165°. The sign and magnitude patterns are computed using the neighborhood pixel values in all directions. The sign pattern provides the local information of an image and is computed by comparing the neighborhood pixels. The magnitude pattern also provides additional information on images. The performance of the proposed method is compared with other existing methods by conducting experiments on five benchmark databases namely Brodatz, Outex, CUReT, UIUC, and Virus. The experimental results prove that the performance of the proposed method has achieved higher retrieval rate than other existing approaches.