Stochastic Systems Group
Home Research Group Members Programs  
Demos Calendar Publications Mission Statement Alumni

SSG Seminar Abstract


An Information Theoretic Approach to Image Segmentation

Junmo Kim
SSG Doctoral Student


In this talk, I will present a novel information theoretic approach to image segmentation. In the proposed method, we segment a given image into the foreground and the background by evolving a closed curve so that we maximize the mutual information between the binary (foreground region inside the curve/ background region outside the curve) label determined by the curve and the image pixel intensity. The evolution of the curve in the gradient direction is based on nonparametric statistics of the regions inside and outside of the curve. I will give some preliminary results and discuss future research avenues.



Problems with this site should be emailed to jonesb@mit.edu