Region-Based Conditional Random Fields For Medical Image Labeling

Abstract

Concerning the high time complexity of medical image labeling in graph model, we proposed a region-based CRF method for medical image labeling.This method first over segmented the image into small homogeneous regions by using over-segmented method, and then the graphical model was constructed with regions as nodes and connecting the neighboring nodes as edges.The corresponding definition of region-based CRF were proposed and implemented.The experimental results shows that better medical image labeling results are obtained by the region-based CRF model.At the same time, running time is largely reduced, efficiency is improved.