Motion-Compensated Transform Coding

Television signals, which are generated by scanning a scene 30 times a second, contain a significant amount of frame-to-frame redundancy. A large part of this redundancy can be removed by the technique of conditional replenishment.1 r' In conditional replenishment, each frame 1703 is segmented into two parts: background, which consists of picture elements (pels) having intensities similar to the previous frame pels, and moving area, which consists of pels that differ significantly from the previous frame pels. Information is transmitted only about the moving area in the form of prediction errors and addresses of the moving area pels. Conditional replenishment schemes can be improved by estimating the displacement of objects in the scene and using the displacement estimate for predictive coding by taking differences of elements in the moving area with respect to appropriately displaced elements in the previous frame. Such schemes have been referred to as motion-compensated coding schemes.'1-11 Transform domain methods have been widely discussed for bandwidth compression of still images or single frames.12 They can also be used for coding of sequences of television frames by taking a twodimensional spatial transform followed by predictive coding using corresponding coefficients from the spatial transform of the previous frame.13"16 This type of hybrid coding17 relieves the storage problems associated with the use of three-dimensional transform blocks. Such a scheme can be made more efficient for scenes containing objects in motion by using, for prediction, coefficients of blocks from the previous frame that are spatially displaced from the present frame block by an amount equal to the displacement of objects.