A Fast Hierarchical Multilevel Image Segmentation Method using Unbiased Estimators

TL;DR

This paper introduces a fast, hierarchical multilevel image segmentation method that optimizes unbiased estimators of variances, offering an efficient global approach that outperforms some existing techniques.

Contribution

It presents a novel, recursive variance optimization technique for hierarchical image segmentation that is faster and more effective than previous methods.

Findings

The method is faster than traditional segmentation techniques.

It achieves competitive or superior segmentation quality.

The recursive variance relation accelerates the segmentation process.

Abstract

This paper proposes a novel method for segmentation of images by hierarchical multilevel thresholding. The method is global, agglomerative in nature and disregards pixel locations. It involves the optimization of the ratio of the unbiased estimators of within class to between class variances. We obtain a recursive relation at each step for the variances which expedites the process. The efficacy of the method is shown in a comparison with some well-known methods.