Computing optical flow from bio-inspired spherical retina

While the conventional image plane is planar, recently bio-inspired cameras are developed in which the photodetectors are positioned on a hemisphere. Image captured by such a spherical retina camera is a discrete spherical image with irregular pixels. In this paper we propose a novel method of computing optical flow from such discrete spherical images. We simulate a spherical retina as a spherical centroidal Voronoi tessellation image, and also investigate the influence of approximating the spherical pixels as regular ones on the computational accuracy. This paper can be regarded as a preliminary work of spherical retina image processing which is necessary in the near future.