Nonparametric method for extraction of region of interest based on information theory

摘要

We propose a nonparametric method for extraction of region of interest (ROI) based on information theory. A polygonal active contour, whose energy function is defined by Jensen-Shannon divergence, is used to drive the curve to match the boundaries of the ROI. Then, our method is able to solve problems involving arbitrary probability densities for the region intensity. We use this method to extract the ROI of synthesized, aerial and natural images and compare with classical statistical snake. Experimental results demonstrate the power of the proposed extraction method for ROI.