A Human Visual-System Based Block Classification Algorithm For Image Sequence Coders

ABSTRACTFor a long time, most of the hybrid codecs for moving images have been of a block based type. Each block isthen coded rather independently by spending bits according to its statistically relevant contents. But this does notyet mean that the data is visually relevant. To overcome this, sometimes people incorporate some visual parametersinside their coding process. The visual weighting matrix so often used in DCT coding is such an example. However,this approach often results in complications, since the coding algorithm as such is almost never suitable to extractor interpret visual relevance. For instance, DCT coefficients do not represent a visual phenomenon. The problemcan be dealt with more thoroughly by completely separating the process into a visual analysis step and a codingstep. An analysis step can now assign attributes to visual features, irrespectively of the actual coding algorithm. Itsresults can then be used by the coding engine to chose a suitable algorithm, to adapt parameters or to chose a localquality criterion. In this way, it is possible to code better adapted to local visual features. This paper will describea codec using such a classification algorithm and present some results.1. INTRODUCTIONAlmost all image compression tools are based on divide and conquer techniques: images are subdivided into