Region matching for pre-operative and post-operative brain images

The problem addressed in this paper is matching corresponding regions in two images, even when the image has correspondence deficiency and local deformations. We present a novel algorithm for establishing region correspondences across images by graph matching and a novel local histogram based feature descriptor. Firstly, we segment the images into structural meaningful regions using simple linear iterative clustering (SLIC). Secondly, we extract the features of the regions and combine the most salient features into an attribute vector for the region. Thirdly, we use the spectral graph matching to compute the correspondence of the regions. The experimental results show that the algorithm has a good performance.