Context partitions in lossless image compression: an enumerative experiment

A context-based lossless image compression coder is used in preference to a first-order entropy coder in the expectation that it provides lower compression ratio. Here we study the potential of improvement in context-based lossless image compression coder underlying the multiresolution method. In multiresolution image compression, an image is decomposed into a set of wavelet coefficients, then the wavelet coefficients are transformed to a set of samples using the alphabet and sample-set partitioned. After sample transformation, inter-subband and intra-subband contexts are used for encoding each sample in the subbands. Specifically, all partitions of inter-subband and intrasubband contexts derived from three reversible wavelet transforms, are enumerated and one first-order and two higher-order entropy encoding coders are applied on the derived samples. We look at the compression ratio as the function of the number of context partitions of intra-subband and inter-subband contexts; the chances of the improvement if we use a different scanning scheme; and the coder schemes performance.