Block-structured interpolative coding of image sequences based on noncausal models

With the popularity of transform-based coding algorithms and the fact that they can now be implemented at real-time video sampling rates, interpolative coding techniques are thought to be worthy of re-examination. The convenience of a block-based coding algorithm should not be overlooked, it can provide the basis for: parallel processing, block concealment in the event that channel errors are detected, and by transmission of a few overhead bits per block, can permit blockwise adaptation schemes, such as motion-compensation to be incorporated. The authors review some of the basic theory, present a study of the quantisation noise amplification of interpolative coding including the design of a MMSE quantisation noise compensator, discuss the transmission of the boundary conditions, and propose a complete coding scheme which is illustrated by numerical example