A robust motion detection technique using unit gradient vectors

This paper presents a robust motion detection technique using unit gradient vectors (UGVs). UGVs carry only gradient orientation information that is remarkably invariant to changes of image intensities. Hence, UGVs can be used as a stable textural feature in place of conventional image features intensities and gradients. We propose to detect motions in image sequences using the pixel-wise variations of the UGVs over time. Simulation results show that the proposed method performs motion detection robustly irrespective of varying image intensities. The method also works well for low-contrast images.