An intelliegent video noise reduction method using adaptive spatial and motion-compensation temporal filter

In this paper, we propose an effective noise reduction method for image sequences corrupted by the Gaussian noise or impulse noise. The basic strategy is to combine the spatial just noticeable distortion (JND) with local image characteristics for spatial filtering and utilize the motion compensation for temporal filtering. For spatial filtering, an adaptive scheme composed of the harmonic mean filter, weighted arithmetic mean filter, alpha-trimmed mean filter, median filter and thresholding filter is dedicated to reducing noises on an image. Then, a motion-compensation based temporal filter is focused on refining the spatial-filtered image frame with the previous and following frames. Experimental results show that the proposed noise-reduction method is better than four previous methods with a PSNR improvement rate of 8.85% on Gaussian noise, 11.69% on fixed-value impulse noise and 11.64% on random-value impulse noise over the average of these four methods.