Optical flow estimation from multichannel spherical image decomposition

The problem of optical flow estimation is largely discussed in computer vision domain for perspective images. It was also proven that, in terms of optical flow analysis from these images, we have difficulty distinguishing between some motion fields obtained with little camera motion. The omnidirectional cameras provided images with large filed of view. These images contain global information about motion and allow to remove the ambiguity present in perspective case. Nevertheless, these images contain significant radial distortions that is necessary to take into account when treating these images to estimate the motion. In this paper, we shall describe new way to compute efficient optical flow for several camera motions given synthetic and real omnidirectional images. Our formulation of optical flow estimation problem will be given in the spherical domain. The omnidirectional images will be mapped on the sphere and used in multichannel image decomposition process to constraint spherical optical flow equation. This decomposition is based on spherical wavelets. The optical flow fields obtained using our proposed approach are illustrated and compared with multichannel image decomposition method developed for perspective images and other published methods dedicated to omnidirectional images.

► We propose multichannel spherical image decomposition method to compute optical flow from omnidirectional images.
► This problem is reformulated on the unit sphere using spherical wavelets.
► Experimental results show that our method is more robust and gives good results compared to differential methods.