| Video down-conversion decoding is a practical and attractive technology, which discard the useless information in the decode loop to reduce the cost of computation or memory.First, this paper introduces the the video compression related concepts and the basic idea of down-conversion decoding. On this basis, this paper focuses on two approaches of down-conversion decoding: spatial down-conversion decoding and temporal down-conversion decoding. This paper introduces a frequency-splitted model for motion-compensation process. Based on the model, the source of drift problem in spatial down-conversion decoding is found and we propose an spatial down-conversion dedoder with high-ferquency motion compensation. The PSNR gain over tranditional spatial down-conversion decoder is about 1.5dB - 3dB, based on different simplify model.For temporal down-conversion decoder, this paper focuses on the acceleration of decoding the dropped P-frame, introducing jump motion compensation(JMC) and proposing a temporal down-conversion decoder model based on JMC. Then the influence of high frequency DCT coefficients on the quality of decoded picture is discussed, and we proposed a acceleration method by cutting some high frequency DCT coefficients.This paper also focuses on the related DCT domain decoder, analyses the rounding mismatch between DCT domain decoder and spatial domain decoder. By the observation of different mismatch distribution in smooth and rough block, a block-division-based compensation algorithm is proposed to compensate the interpolation rounding mismatch. It can improve the PSNR by 0.3dB. Also, in order to compensate the IDCT rounding mismatch, this paper proposes a DC-block-based compensation algorithm. |
PDF Full Text Request