Robust Optical Flow Computation Under Varying Illumination Using Rank Transform

Variational method is regarded as one of the best performing optical flow techniques, which determine the optical flow field by minimizing a suitable energy function consisting of a data term and a regularization term. The brightness constancy assumption has widely been used in the data term, which is obviously violated to the real case. To improve the ability of optical flow’s application in outdoor computer vision task, a morphologically invariant descriptor called rank transform is introduced and the tolerance to illumination changing is analysis in this paper. Due to its good performance under illumination morphological change, we embed it into the data term of TV_L1 model to get robust optical flow field. Experiments with Middlebury and KITTI benchmark demonstrate the good performance of method proposed. Copyright © 2014 IFSA Publishing, S. L.