Evaluation of Alpha-Trees for Hierarchical Segmentation by Horizontal Cuts

Area of Science:

• Computer Vision
• Image Processing
• Data Structures

Background:

• Alpha trees and alpha-omega hierarchies are established methods for hierarchical image representation.
• The quality assessment of these hierarchies is underdeveloped, hindering their broader application and optimization.

Purpose of the Study:

• To propose a novel algorithm for evaluating the quality of alpha-omega hierarchies.
• To enable automatic selection of optimal parameters and dissimilarity measures for hierarchy construction.
• To address limitations in current hierarchical image representation techniques.

Main Methods:

• Development of an evaluation algorithm based on horizontal cut filters.
• Systematic consideration of factors like accuracy, complexity, and efficiency in hierarchy construction.
• Experimental validation using remote sensing images.

Main Results:

• The proposed algorithm effectively evaluates alpha-omega hierarchy quality.
• Demonstrated usefulness of the algorithm through experiments on remote sensing data.
• The algorithm facilitates automatic selection of optimal construction parameters.

Conclusions:

• The developed algorithm provides a robust method for assessing alpha-omega hierarchy quality.
• This work advances the field of hierarchical image representation by enabling better parameter selection.
• The algorithm's potential extension to other hierarchical tree types broadens its applicability in image segmentation.