A Fast Hierarchical Multilevel Image Segmentation Method using Unbiased Estimators
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.
cs.CV↗