Multidimensional signal compression using multi-scale recurrent patterns with smooth side-match criterion

The recently proposed method for image compression based on multi-scale recurrent patterns, the MMP (multidimensional multiscale parser) has been shown to perform well for a large class of images, specially for those containing text or graphics. However, its performance for coding smooth, gray scale images is still distant from the state of the art. In this paper we propose an extension for it, the SM-MMP (side-match MMP). In this method, as in MMP, a multidimensional signal is recursively segmented into variable-length blocks, and each segment is encoded using expansions and contractions of vectors in a dictionary. The dictionary is updated while the data is being encoded, using concatenations of expanded and contracted versions of previously encoded blocks. However, unlike MMP, in SM-MMP the dictionaries are built considering smoothness constraints around block boundaries, similar to the side-match vector quantization methods. This allows it to perform better than the MMP when the images are smooth, without sacrificing its performance for images containing text or graphics. Indeed, our simulation results show that the proposed method is effective, yielding improvements of the order of 1.5 dB over the original MMP for grayscale images, while preserving the high performance of the original MMP for graphics, text and mixed images.