A phase-based framework for optical flow estimation on omnidirectional images

Over the past few years, omnidirectional vision has become an important area of research because omnidirectional cameras offer a large field of view compared with conventional perspectives cameras. However, omnidirectional images contain important distortions, and classical optical flow estimations are thus not appropriate. In this paper, we propose to estimate optical flow on omnidirectional images using a phase-based method which proved its robustness and its accuracy on the perspective images. We will adapt different treatments in order to take into account the nature of omnidirectional images.