An Efficient Re-quantization Error Compensation for MPEG2 to H.264 Transcoding

During transcoding, the coefficients have to pass through another quantization step. This introduces re-quantization errors to the coefficients. H.264 integer transform and quantization features are different from those of MPEG2 and other standards. Based on these features, in MPEG2 to H.264 transcoding, an efficient algorithm that measures the re-quantization error is proposed. Then this measured error is used to compensate for the quality loss in transcoding. The experimental results from four typical video test sequences show that the proposed compensation procedure improves the PSNR value by about 4.82 dB. The error calculation and compensation could be carried in the transform domain resulting in significant computational saving