Motion compensated enhancement of noisy image sequences

A motion compensated image sequence enhancement algorithm is presented. A combined segmentation and motion estimation algorithm is employed. A temporal or a spatiotemporal low-pass filter is then applied. Mean and median filters are presented as low-pass filters. The temporal filtering is performed over the motion path of each pixel, which is provided by the motion-estimation algorithm. The spatial filtering does not blur the boundaries of the moving objects because the boundary locations are provided by the segmentation algorithm. The performance of the combined algorithm is examined using computer-generated and real image sequences corrupted by additive white Gaussian noise. The algorithm performs very well in a very noisy environment. Mean filtering is more effective in the case of white Gaussian noise, and median filtering is more effective in the case of salt-and-pepper noise and burst noise