TECHNIQUES FOR THE INCREASE OF PERFORMANCE OF FORWARD TRANSFORMS AND MOTION ESTIMATION DEDICATED ARCHITECTURES OF THE H.264/AVC VIDEO CODING STANDARD / TÉCNICAS PARA O AUMENTO DE DESEMPENHO DE ARQUITETURAS DEDICADAS DAS TRANSFORMADAS DIRETAS E DE ESTIMAÇÃO DE MOVIMENTO DO PADRÃO H.264/AVC DE CODIFICAÇÃO DE VÍDEO PELOTAS 2009

AUTOR(ES)

ANDRÉ MARCELO COELHO DA SILVA

DATA DE PUBLICAÇÃO

2009

RESUMO

The increasing use of digital video on the Internet, devices and also in mobile digital TV has lead to an increasing interest for research in this area, from both industry and academia. This work follows this trend by proposing the investigation of the main aspects of digital video, video compression and also of the H.264/AVC video compression for the implementation of performance efficient dedicated architectures for some modules of the H.264/AVC encoder. In particular, this work presents some architectural alternatives for the increase of performance of two modules of the H.264/AVC, which are: T Module (composed by the Forward Hadamard and Discrete Cosine Transforms) and Motion Estimation (ME). For the implementation of these modules the use of techniques to the increase of performance has been considered, such as the use of efficient adders and pipeline. This efficient adders presents high use of hardware features. Thus, the implemented architectures in this work presented these characteristics too. The main aspect presented by the implemented modules is a large number of arithmetic operations of addition and subtraction for their processing. Thus, the motivation of this work is the increase of performance of these modules, from the use of efficient adder/subtractor circuits that are present in literature. In particular, 4:2, 8:2 and 16:2 adder compressors that perform the simultaneous addition of 4, 8 and 16 operands, respectively, with no penalties in area and the critical path are used. The architectures were described in VHDL and targeted to ASIC technology. The validation of the circuits and the obtained results were performed by using Leonardo Spectrum tool from Mentor Graphics. Comparisons against the solutions of the literature were done and the main results show that the architectures proposed in this work are more efficient. Significant gains in performance are achievable using our solutions for both Forward Transforms and Motion Estimation architectures

ASSUNTO(S)

estimação de movimento padrão h.264/avc h.264/avc standard compressão de vídeo motion estimation vídeo digital ciencia da computacao video compression forward transforms transformadas diretas digital video

Documentos Relacionados

• Desenvolvimento de Arquiteturas de Alto Desempenho dedicadas à compressão de vídeo segundo o Padrão H.264/AVC
• Modelagem de hardware para codificação de vídeo e arquitetura de compensação de movimento segundo o padrão H.264/AVC
• Implementação física de arquiteturas de hardware para a decodificação de vídeo digital segundo o padrão H.264/AVC
• Desenvolvimento arquitetural para estimação de movimento de blocos de tamanhos variáveis segundo padrão H.264/AVC de compressão de vídeo digital
• Arquitetura de hardware dedicada para a predição intra-quadro em codificadores do padrão H.264/AVC de compressão de vídeo