Detecting, grouping, and structure inference for invariant repetitive patterns in images

Area of Science:

• Computer Vision
• Image Analysis
• Pattern Recognition

Background:

• Invariant pattern extraction is a core challenge in computer vision.
• Repetitive patterns are key to understanding image textures, object shapes, and composition.
• Existing methods struggle with robust and efficient detection of these patterns.

Purpose of the Study:

• To develop a new algorithm for detecting and grouping repetitive patterns in images.
• To infer the composition structure of detected repetitive patterns.
• To leverage pattern grouping for inferring object geometry and scene layout.

Main Methods:

• A region growing segmentation scheme is employed.
• A mean-shift-like dynamic is used for clustering local image patches.
• Continuous joint alignment is utilized for patch matching and subspace grouping refinement.

Main Results:

• The algorithm successfully detects and groups repetitive patterns.
• A novel inference algorithm constructs a structural completion field for global geometric structures.
• The grouping results aid in inferring object geometry and scene layout.

Conclusions:

• The proposed algorithm offers an efficient and robust approach to repetitive pattern extraction.
• The method effectively merges detected patterns into global geometric structures.
• This work advances the understanding of image interpretation for complex scenes.