Recursive Motion Compensation: A Review

TLDR: Algorithms for estimating frame-to-frame changes in intensity of a television scene, for motion compensated interframe coding, adaptive predictors and a segmenter to determine which pels need to be transmitted are presented.

Abstract: We present algorithms for estimating frame-to-frame changes in intensity of a television scene. The changes can be a result of object motion in a TV scene or variation of illumination. By a gradient search technique, which seeks to minimize a functional of the interframe intensity prediction error, we estimate two parameters associated with these intensity changes — translational displacement of objects and a scaling factor associated with illumination changes. Algorithms for the parameter estimation are given in both the picture element and transform domain. Using the estimates of these parameters, we also develop, for motion compensated interframe coding, adaptive predictors and a segmenter to determine which pels need to be transmitted. We describe several coder variations and compare them by computer simulations using several scene sequences. Simulations indicate that motion compensation reduces the bit rates by 30 to 70 percent compared to conditional replenishment.