A Graph Based Image Interpretation Method Using A Priori Qualitative Inclusion and Photometric Relationships.

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Image segmentation often produces oversegmentations.
• Existing methods rely heavily on spatial and quantitative data.
• Qualitative information is underutilized in region recovery.

Purpose of the Study:

• To develop a method for recovering and identifying image regions from oversegmentations.
• To leverage qualitative knowledge, including inclusion and photometric relationships.
• To offer an alternative to purely quantitative and spatial approaches.

Main Methods:

• A two-step inexact graph matching approach is proposed.
• Subgraph isomorphism is used to match expected regions with oversegmented subsets.
• Progressive merging of remaining regions preserves qualitative relationships.

Main Results:

• The method successfully recovers expected image regions described qualitatively.
• It demonstrates effectiveness across various images (grayscale, color) and oversegmentation algorithms.
• Performance is evaluated on a custom experimental database.

Conclusions:

• The proposed method offers a viable approach for image region recovery using qualitative information.
• It achieves recovery in reasonable runtimes, providing an open-source Python package for accessibility.
• This work highlights the potential of qualitative reasoning in image analysis.