Improved Segmentation Based on Probabilistic Labeling

Abstract

This paper presents an improved multi-object segmentation algorithm based on probabilistic labeling. First, a critical look is focused on utilizing vector calculus operator and combinational operator to rewrite Dirichlet integral into a matrix form, and boundary condition is defined to obtain the needed harmonic function. The only unique parameterβthat dominantly affects the segmentation performance is characterized. According to the result, we propose an improved parameter that changes the value ofβon the basis of pixel-by-pixel, rather than the use of a fixed constantβthroughout the whole image. Furthermore, a pre-process involving the use of watershed analysis is applied to smooth the effect of high frequency components in the input image, so that better noise tolerance and more accurate object contours can be obtained.