Image Predictive Coding using Multiscale Recurrent Patterns

The multidimensional multiscale parser (MMP) algorithm was originally proposed as a multidimensional signal lossy compression method. This method uses a set of patterns stored in a dictionary, that are scaled to approximate each variable-length segment of a multidimensional signal. Thus, the input signal is encoded using only a set of flags, that define the segmentation of the original message, and the dictionary indexes that encode each of these vectors. MMP has proved to be very effective for encoding image data, achieving, for certain classes of images, like texts and graphics, relevant gains over state-of-the-art algorithms, like SPIHT and