Codificação de imagem e vídeo com casamento parcial de padrões em transformadas de bloco.

Abstract

This doctoral dissertation aims to develop and evaluate an efficient technique. for coding without loss of quantized coefficients from the application block transforms, applied to image and video compression. To this end, we employ Partial Pattern Matching (CPP) techniques such as pre-processing of the entropic coding step of transform-based image and video encoders. For image we use the already coded blocks as a search dictionary. At the In this video case, we analyze the compression in the residue images obtained by the estimation and motion compensation using previous residue images in the formation of the dictionary. The partial pattern matching algorithm is based on the fact residual correlation between the significance maps of the coefficient blocks transformed. Similar blocks serve as a reference for coding by providing a map with probabilities for encoding the obtained residue. For the implementation From these experiments, modifications were made to an existing MPEG-4 encoder to inclusion of these algorithms to perform experiments to analyze the merits of the technical as well as objective results. Some factors like dictionary types, search methods and metrics to be used in defining the distance to be minimized are discussed. Here we present some contributions to the problem of coefficient coding for image and video compression.