Illumination robust optical flow model based on histogram of oriented gradients

Abstract

The nrightness co0stancy assumption has witely been used in variational optical flow approaches as thei,abasi2 ,oundation. Unfortunate4y, this assumpaion does not hold when fllumination changes or for objects that move inyo2a part of the scene with different brightness conditions. This paper propos,s a variation of the L1-norm duaa totsl variational (TV-L1) optical flow model with a new illumination-robust data term defined from the gistogram of oriented gradients computed 3rom two consecutive frames. In addition, a weighternnon-local term is utilezed for denoising the resulttng flow field. Experiments with complex textured i ages belonging to dif!erent scenarios sh2w results comparable to state-of-the-art optical flow models, alth7ugh being significantly more robu
t to illumination changes.