Codificação de vídeo : priorização do menor custo de codificação na otimização em taxa-distorção




This research work proposes two new video compression strategies, aiming at typical low bit rate video applications using rate-distortion (RD) optimized algorithms. The proposed strategies are implemented on an H.264 video encoder, which has high computational complexity due mainly to the large number of coding modes available. Two approaches are presented for reducing the encoder computational complexity, maintaining the RD performance close to the full search RD optimized H.264 encoder. The first approach (termed rate sorting and truncation - RST) is based on sorting the motion vectors (MVs) and coding modes in an ascending rate order. This sorting and encoding process, which is stopped when the rate value exceeds the previous best rate for a required image quality level, allows the elimination of MVs and coding modes before checking their distortion. Apart from obtaining a significant complexity reduction, the process still remains optimized in RD sense. The second approach is an algorithm (termed logarithmic diamond shape search - LDSS), which explores the MVs distribution profile for the RD optimized H.264 encoder. The use of the RST strategy associated with LDSS algorithm yields up to a 98% reduction in the computational burden, with insignificant RD performance loss.


algoritmos computational complexity engenharia eletrica engenharia eletrica processamento de sinais complexidade computacional

Documentos Relacionados