Fuzzy Color Aura Matrices for Texture Image Segmentation.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Fuzzy gray-level aura matrices are effective for color texture classification but computationally intensive for segmentation.
• Existing methods struggle with high memory and computation requirements for color texture segmentation.

Purpose of the Study:

• To extend fuzzy gray-level aura matrices to fuzzy color aura matrices for improved color texture image segmentation.
• To develop an adaptive neighborhood function for enhanced segmentation accuracy.
• To reduce computational complexity in color texture segmentation.

Main Methods:

• Extension of fuzzy gray-level aura matrices to fuzzy color aura matrices to capture inter-color pixel interactions.
• Development of an adaptive neighborhood function using a modified Simple Linear Iterative Clustering (SLIC) algorithm for superpixel generation.
• Superpixel classification using fuzzy color aura matrices as features with a supervised classifier.

Main Results:

• The proposed fuzzy color aura matrices effectively characterize local color interactions between neighboring pixels.
• The adaptive neighborhood function improves segmentation by incorporating regional information.
• Experimental results demonstrate superior performance compared to classical supervised methods and comparable results to deep learning methods on the Prague texture segmentation benchmark.

Conclusions:

• Fuzzy color aura matrices offer a computationally efficient and accurate solution for color texture image segmentation.
• The adaptive neighborhood approach enhances segmentation robustness.
• The method achieves state-of-the-art performance, rivaling deep learning techniques.