Two-Dimensional EspEn: A New Approach to Analyze Image Texture by Irregularity

Area of Science:

• Image Processing
• Computer Vision
• Data Science

Background:

• Texture analysis is crucial for image classification and segmentation.
• Existing entropy algorithms like Shannon Entropy and SampEn2D have limitations in capturing spatial information and reliability in 2D data.
• There is a need for advanced algorithms to quantify irregularity in 2D data.

Purpose of the Study:

• Introduce Espinosa Entropy (EspEn), a novel algorithm for measuring irregularity in 2D data.
• Evaluate EspEn's performance against established entropy methods.
• Assess EspEn's sensitivity to noise and its applicability to texture analysis.

Main Methods:

• Developed EspEn algorithm with parameters: m (window length), r (tolerance threshold), and ρ (similarity percentage).
• Conducted experiments on simulated images with varying noise levels and grayscale images from the Normalized Brodatz Texture (NBT) database.
• Compared EspEn with Shannon Entropy and SampEn2D, and analyzed parameter influence on EspEn's performance.

Main Results:

• EspEn demonstrated superior ability to discriminate images based on size and noise levels.
• The algorithm showed reliable performance in quantifying image irregularity.
• Recommended parameters for optimal EspEn performance are m=3, r=20, and ρ=0.7.

Conclusions:

• EspEn offers a robust alternative for quantifying irregularity in 2D data, particularly in image processing.
• The algorithm effectively addresses limitations of existing entropy-based methods.
• EspEn shows promise for applications in texture analysis, image classification, and segmentation.