This paper presents a new algorithm focusing in high quality fast motion estimation for high definition video coding. This algorithm provides more efficiency to avoid local minima falls in fast motion estimation due to random search exploration. The developed algorithm is called Random Search with Iterative Center Evaluation and Iterative Final Refinement (RSICIFR). It evaluates the central region of the search area, and also chooses, randomly, many other candidate blocks in the search area. This approach gives to the algorithm the possibility to avoid local minima falls, increasing the quality results, especially for high definition videos. The RSICIFR algorithm was implemented and evaluated to ten HD 1080p test video sequences. The evaluation results shows that a quality gain can be achieved in comparison of many fast algorithms.
http://www.lbd.dcc.ufmg.br/colecoes/sim/2011/0020.pdfCaso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web