Medical image registration based on local feature of topological texture image and optical flow model
YU Wenyon · Journal of Computer Applications · 2014
Medical image is a kind of topological texture image. In clinical diagnosis and treatment, it is often needed to match medical images of different models. This paper put forward a medical imaging optical flow field registration method in conjunction with invariant features of topology texture image. Firstly, Scale Invariant Feature Transform( SIFT) descriptor was used to describe local characteristics of image registration and image registration on local imaging characteristics is refined by the consistency purification method. Secondly, the image was matched on the basis of elastic image registration of optical flow model by reference to the image local invariant features, and the registration model was decomposed by units to avoid blurred registered result and loss of local topological image texture features because of being smoothed. The result of simulation experiment indicates that the new method has better effect on medical image registration.