Texture Analysis And Characterization Using Probability Fractal Descriptors

TL;DR

This paper introduces a novel texture analysis method using probability-based fractal dimension estimation and multiscale transforms, demonstrating high effectiveness in texture classification tasks.

Contribution

It presents a new fractal dimension-based texture descriptor using the probability (Voss) method combined with multiscale analysis, improving texture characterization.

Findings

High classification accuracy on benchmark datasets

Effective multiscale transform enhances texture discrimination

Proven utility for texture analysis and characterization

Abstract

A gray-level image texture descriptors based on fractal dimension estimation is proposed in this work. The proposed method estimates the fractal dimension using probability (Voss) method. The descriptors are computed applying a multiscale transform to the fractal dimension curves of the texture image. The proposed texture descriptor method is evaluated in a classification task of well known benchmark texture datasets. The results show the great performance of the proposed method as a tool for texture images analysis and characterization.