An Adapted Block-Matching method for optical flow estimation in catadioptric images

Optical flow estimation is one of the basic problems in computer vision. It is largely discussed in the case of perspective images. The omnidirectional cameras provide images with large field of view. These images contain global information about motion and permit to remove the ambiguity present with little camera motion in perspective case. Nevertheless, these images contain significant radial distortions. Optical flow estimation of motion in these catadioptric cameras is a complex problem and still a challenging research field. In this paper, we shall present new a Block-Matching approach to efficiently compute optical flows for different catadioptric camera motions. Based on an adapted neighborhood, the motion field allows accurate estimation of motion in catadioptric images. The optical flow fields obtained using our proposed approach are illustrated on synthetical and real images and compared with method developed for perspective images.