On transform coding of motion-compensated difference images

Motion-compensated interframe transform coding has been recommended as an efficient method of coding image sequences at low bit rates. The author considers the effect of employing a variety of different transform and motion compensation techniques on the operation of the algorithm and also the interrelation between the predictive and transform parts of the scheme in terms of the spectral properties of the signal to be transformed. It is shown that, for a wide variety of source material, there is a complex interaction between the specific transform and motion compensation scheme employed, and that the efficiency of the motion-compensated predictive operation strongly influences the performance of the subsequent transform step.<>