Patent US 5,544,239

A method and apparatus for improving motion analysis in fade regions where motion compensation systems based on block matching typically fail to accurately encode images. A previous image is selected from an image sequence. A base image is calculated by adjusting the brightness of the previous image. A displacement vector representing the magnitude and direction of the displacement between at least one region in a target image and a corresponding region in the base image is determined. The displacement vector is applied to a corresponding region in a previous reconstructed image to form a predicted image, and is then encoded. Pixel values in the predicted image are subtracted from corresponding pixel values in the target image to form an error image which is also encoded.