Robust image registration in the gradient domain

In many real-world applications of image registration, the images have significantly different appearances due to the intensity variations. Many existing intensity based methods may fail to solve these challenging problems. In this paper, we propose a novel method based on the differential total variation (DTV) for image registration. It is inspired by the fact that the image gradients are much more stationary than the intensities, especially when there exist severe intensity distortions. Therefore, we prefer to register the images in the gradient domain, which intuitively leads to more accurate registration results. An efficient algorithm is presented to solve the DTV minimization problem. The proposed method is scalable and has no regularization parameter to be tuned, both of which are desired properties for image registration. We show the accuracy and efficiency of our method through extensive non-rigid registration experiments, on synthetic MR images and real retina and iris images.