DCT Block Conversion for H.264/AVC Video Transcoding

In H.264/AVC [1], integer transforms are applied instead of the 8×8 discrete cosine transform (DCT) of previous standards to avoid inverse transform mismatch problems. However, these transforms make it difficult to transcode the precoded video contents with the earlier video coding standards into H.264/AVC in DCT-domain, thus causing cascaded re-encoding in the pixel-domain. There are strong requests for an efficient way to solve this problem. In this paper, we propose a new conversion scheme from an 8×8 DCT block into four 4×4 DCT blocks in H.264/AVC. Experimental results show that the proposed scheme improves PSNR up to 0.1dB and a little computational complexity compared with the cascaded pixel-domain IDCT/DCT. Moreover, other DCT-domain transcoding methods are applicable based on our scheme.