Feature points tracking : robustness to specular highlights and lighting changes

Since the accurate modelling of relection is a difficult task, most feature points trackers assume that objects are lambertian and that no lighting change occurs. To some extent, a few approaches answer partially these issues by computing an affine photometric model or by achieving a photometric normalization. Through a study based on specular reflection model (Phong and Torrance-Sparrow models), we explain explicitly the assumptions on which these techniques are based. Then we propose a tracking method that compensates for specular highlights and lighting variations more efficiently when small windows of interest are considered. It supposes that the illumination variations can be approximated by a continuous function around the point to be tracked. Contrary to the existing approaches, the illumination variations are not supposed to be constant in each point of the window of interest. experimental results on different image sequences prove the robustness and the accuracy of this technique in comparison with the existing trackers. Moreover, the computation time of the tracking is not significantly increased.