HDTV picture quality performance in the presence of random errors, analysis and measures for improvement
作者:
Highlights:
•
摘要
We study the performance of the picture quality of HDTV in the presence of random errors and find the different kinds of picture impairments that may arise. The problems associated with picture coding, including variable length and Huffman coding, are investigated in great detail. The use of DC differential coding and the loss of block synchronization cause errors to perpetuate to other regions of the picture, and hence are identified as the major causes of poor performance. Motivated by these observations, two novel techniques which significantly improve the picture quality are proposed, analyzed and evaluated. The first technique is based upon encoding the actual values of the DC coefficients instead of the commonly used differential DC coding. This is done by means of a new set of Huffman codewords. It is shown that, contrary to present belief, by coding the actual DC values it is possible to maintain the high compression obtained by the commonly employed differential DC coding and at the same time significantly improve the noise resistance of the coding scheme. The second proposed technique introduces a simple but effective synchronization scheme which restricts the effects of residual errors to within 32 × 16 pixel superblocks thus improving the quality of the received picture. This synchronization area is much smaller than the 5632, 4024 and 3072 pixel areas proposed by the four HDTV proposals. Incorporation of the two techniques yields a scheme which has excellent noise performance characteristics and increases the overall noise resistance of the HDTV system. This scheme is applied to both reference frames and inter-frames and improves the SNR at which the HDTV picture suddenly deteriorates by 2.5–3 dB. After that the deterioration is more graceful, as the error propagation is limited to the superblock boundaries.
论文关键词:
论文评审过程:Received 12 August 1994, Available online 12 February 1999.
论文官网地址:https://doi.org/10.1016/0923-5965(95)00037-2