In this work, a multigrid motion compensation video coder scheme based on the current HVS contrast discrimination models is proposed. A novel procedure for the prediction errors encoding, which restricts the maximum perceptual distortion in each transform coefficient, has been developed. This subjective redundancy removal procedure includes the amplitude non-linearities and some temporal features of human perception. A perceptually weighted control of the adaptive motion estimation algorithm has been derived from this model. Perceptual feed-back in motion estimation ensures a perceptual balance between the motion estimation effort and the redundancy removal process. The results show that this feed-back induces a scale-dependent refinement strategy that gives rise to more robust motion estimates which facilitate higher level sequence interpretation. The comparison of the proposed scheme versus a H.263 scheme with unweighted motion refinement and MPEG-like quantization shows the overall subjective superiority of the proposed improvements.
 

References: 47