Occlusion handling in meanshift tracking using adaptive window Normalized Cross Correlation

Meanshift is an efficient tracking technique that has been demonstrated to be robust to small camera motion, clutter and scale changes. Its performance compromises, though, under occlusion and/or complete loss of the object for few frames. In this paper, Normalized Cross Correlation is applied as an additive step for occlusion handling event, initiated by Bhattacharyya Coefficient threshold. Experimental results show that these improvements make meanshift more robust and accurate for handling occlusions and/or complete loss of the object during tracking.