Comparing function or morphology between individuals requires non rigid matching

because the detail spatial structure difference between the image pair to be matched is too complicated to be modeled by any parameterized transformation. The goal of deformable matching method is to remove structural variation between the image pair to be matched. In this paper

a new method of deformable image matching based on hybrid elastic models (HEM) is proposed. The method

which need not extract features

works directly on grey level images. The algorithm first globally aligns images with a principal axis method

and then utilizes the linear spring net model for the correspondence and the thin plate spline for the non rigid mapping. This method takes multiresolution strategy to approach better matching. The elastic constant of the spring model will decreases as the process proceeds. Some experiments are performed on both synthetic and segmented medical images. It is shown that our hybrid elastic models can be successfully applied into the deformable image matching to remove the detail structural variation

and achieve good results.