Color Texture Classification Combining LBP Images and Fractal Features

Abstract

This paper describes a color texture classification scheme that uses fractal features like: box-counting fractal dimension and differential box-counting fractal dimension. Color textured images can be represented on multiple color spaces and recent developments for color texture analysis and classification are increasing. Following this concept, a texture image is obtained using Local Binary Pattern whereupon fractal features are extracted from it. For testing this color texture classification scheme two datasets (BarkTex and VisTex) are used, and the results are compared with three other methods. We propose a modified formula for calculating the Local Binary Pattern of a color textured image and advance the algorithm for this technique. The fractal features prove to be suitable in the classification process and the accuracy rates we had obtained are close to state-of-the-art approaches using a much smaller feature space.