Enhancing motion trajectories in noisy video sequences

An approach for enhancing motion trajectories in video sequences is presented. The proposed approach is applied to projections of image data of video sequences and consists of six enhancement steps. The basic idea is to correlate the positive and the negative derivative part with respect to time to the projected image sequence. Both parts are aligned at their correlation maxima and they are element-wise multiplicated to suppress noise. Morphological operators are used to obtain a one pixel wide motion trajectory.