Comparison of two generating kernels in linked-pyramid segmentation of noisy images

Linked node pyramid segmentation is a cooperative and iterative segmentation process. The first step of carrying out the segmentation is forming an initial pyramid. Overlapped local averaging has been used as the blurring scheme for the formation of the initial pyramid. The paper studies two different blurring filters: Gaussian and morphological filters. The segmentation performance of these filtering schemes are measured according to some objective criteria and the pros and cons of these schemes are discussed