Fast Implementation for Multiscale Recurrent Pattern Image Coding

Multidimensional Multiscale Parser (MMP) algorithms have consistently achieved better compression gains over the years, and recently proposed techniques outperformed state-of-the-art transform-based image encoders like the JPEG2000 and H.264/AVC, for both smooth and nonsmooth images, at low to medium compression rates. The use of dictionary update tools and a highly adaptable segmentation feature allowed the encoder to compress images in a more efficient manner, but the good performance came at the cost of greater computational cost. In this paper we address the algorithm complexity, revise some developments towards reducing the computational burden and propose a new fast implementation, with little decrease in coding gain. Experimental results show that the modified algorithm is still outperforming JPEG2000 and H.264 for smooth and compound images, with a significant gain in coding time.