Segmentation by texture using correlation.

Area of Science:

• Computer Vision
• Image Processing
• Pattern Recognition

Background:

• Texture segmentation is crucial for image analysis.
• Existing methods may struggle with uniform texture identification.
• Hierarchical image representations offer potential for efficient segmentation.

Purpose of the Study:

• To develop and evaluate a novel texture segmentation algorithm.
• To utilize correlation coefficients for identifying uniform texture regions.
• To improve segmentation accuracy through a hierarchical approach and post-processing.

Main Methods:

• Correlation coefficients were evaluated on a quadratic picture tree (pyramid).
• A split-and-merge algorithm was employed for preliminary segmentation based on texture uniformity.
• A grouping algorithm, region adjacency graph, and small region elimination refined the segmentation.

Main Results:

• The split-and-merge approach successfully identified and segmented regions based on texture similarity.
• The grouping algorithm and small region elimination enhanced the final segmentation quality.
• The method demonstrated effective texture segmentation on example images.

Conclusions:

• The proposed method effectively segments images based on texture using correlation coefficients.
• The hierarchical split-and-merge strategy combined with post-processing yields robust texture segmentation.
• This approach offers a reliable way to analyze texture in image processing applications.