RGB - Based color texture image classification using anisotropic diffusion and LDBP

Abstract

In this paper, a novel color texture image classification based on RGB color space using anisotropic diffusion, and local directional binary patterns (LDBP) is introduced. Traditionally, RGB color space is widely used in digital images and hardware. RGB color space is applied to obtain more accurate color statistics for extracting features. According to characteristic of anisotropic diffusion, image is decomposed into cartoon approximation; further the texture approximation is obtained by subtracting the original image and cartoon approximation. Then, texture features of image are obtained by applying LDBP co-occurrence matrix parameters on texture approximation. LDA is used to enhance the class seperability. After feature extraction, k-NN classifier is used to classify texture classes by the extracted features. The proposed method is evaluated on Oulu database. Experimental results demonstrate the proposed method is better and more correct than RGB based color texture image classification methods in the literature. © Springer International Publishing Switzerland 2014.