Texture analysis by multi-resolution fractal descriptors

TL;DR

This paper introduces a multi-resolution fractal-based texture descriptor using Bouligand-Minkowski measures, which improves classification accuracy over traditional methods like Gabor-wavelets and co-occurrence matrices.

Contribution

The paper presents a novel multi-resolution fractal descriptor that recursively analyzes textures and extracts entropy features, outperforming existing texture analysis techniques.

Findings

Achieves higher classification accuracy than classical descriptors.

Effective in texture discrimination on Brodatz and Vistex datasets.

Demonstrates the utility of fractal-based features in texture analysis.

Abstract

This work proposes a texture descriptor based on fractal theory. The method is based on the Bouligand-Minkowski descriptors. We decompose the original image recursively into 4 equal parts. In each recursion step, we estimate the average and the deviation of the Bouligand-Minkowski descriptors computed over each part. Thus, we extract entropy features from both average and deviation. The proposed descriptors are provided by the concatenation of such measures. The method is tested in a classification experiment under well known datasets, that is, Brodatz and Vistex. The results demonstrate that the proposed technique achieves better results than classical and state-of-the-art texture descriptors, such as Gabor-wavelets and co-occurrence matrix.