Augmentation in object tracking performance with hue preservation conditioned on frequency domain characteristics of video

Tracking of any given object forms integral part in surveillance, control and analysis applications. The video tracker presented here works on the principle of mean shift. Mean shift is an iterative algorithm which is extended to the field of object tracking. However mean shift tracker losses track of the object when there are variations in illumination. In order to improve the performance of the tracker the poorly illuminated video frames are pre-processed and enhancement is provided to only those frames based on hue preservation algorithm. Low illumination in a frame leads to low value DC coefficient. Enhancement is provided to frames with low DC coefficient. The frames with corrected illumination are used to track object of interest using the Mean shift algorithm. The proposed method is tried on a variety of video sequences. The results exhibit much improved tracking. The graph plotted for Bhattacharyya coefficient versus frame index exhibits significant increase in its values when the frames are pre-processed in comparison with the basic mean shift algorithm.