Digital video signal encoder and encoding method

Abstract

Motion vectors and associated errors which are determined in the course of encoding a video signal are used to determine whether respective macroblocks of two or more frames represent the same subject matter albeit displaced by motion represented by the frames. The motion vector errors are compared to a predetermined motion threshold. If a particular motion vector error is less than the motion threshold, the macroblocks corresponding to the motion vector error are determined to represent substantially identical subject matter and the corresponding motion vector represents displacement of the subject matter between the frames due to motion of the subject matter. Accordingly, differences between the macroblocks corresponding to the motion vector error are determined to be due primarily to noise and not to differences in the subject. Therefore, a relatively heavy temporal filter is applied between the macroblocks. Conversely, if a particular motion vector error is not less than the motion threshold, the macroblocks corresponding to the motion vector error are determined to represent different subject matter and only a spatial filter is applied to either macroblock. The spatial filter is applied relatively lightly to reduce noise while minimizing signal loss, e.g., blurring, resulting from the filtering. The motion threshold is set according to the noise variance of the video signal which is being encoded. The noise variance of the video signal can be measured using the video signal acquisition equipment used to acquire the video signal, e.g., a video camera, or can be estimated from the video signal itself.