An affine symmetric image model and its applications

Area of Science:

• Computer Vision
• Image Processing
• Computational Mathematics

Background:

• Natural images exhibit significant redundancy, partially addressed by directional wavelets.
• Exploiting geometric redundancy remains a challenge in image analysis.

Purpose of the Study:

• To develop an affine symmetric image model for exploiting geometric redundancy.
• To enable affine invariant texture segmentation and image approximation.

Main Methods:

• A texture model identifies structural patterns.
• Affine symmetry is exploited locally to estimate affine transforms between texture patches.
• The model is extended globally and integrated into a multiresolution framework for multiscale analysis.

Main Results:

• The methodology effectively estimates local and global affine relations within images.
• The multiresolution model successfully exploits image self-similarity across space and scale.
• Experimental evaluation confirms the approach's effectiveness.

Conclusions:

• The affine symmetric image model offers a flexible scheme for exploiting geometric redundancy.
• This approach demonstrates significant potential for practical applications in image analysis, particularly in texture segmentation and image approximation.