Long - Term Memory Prediction Using Affine

Long-term memory prediction extends motion compensation from the previously decoded frame to several past frames with the result of increased coding ee-ciency. In this paper we demonstrate that combining long-term memory prediction with aane motion compensation leads to even higher coding gains. For that, various aane motion parameter sets are estimated between frames in the long-term memory buuer and the current frame. Motion compensation is conducted using standard block matching in the multiple reference frame buuer. The picture reference and the aane motion parameters are transmitted as side information. The technique is embedded into rate-distortion optimal coder control. Signiicant coding gains between 20 and 50 % are achieved for the sequences tested over TMN-10, the test model of H.263. These bit-rate savings correspond to gains in PSNR between 0.8 and 3 dB. As shown in the paper, both parts of the multi-frame prediction scheme, i.e., the long-term memory as well as aane motion compensation, contribute to the overall gain.