Converting DCT Coefficients to H.264/AVC Transform Coefficients

An New Coefficients Transform Matrix for the Transform Domain MPEG-2 TO H.264/AVC Transcoding

“…Relative to Jun Xin's method, our proposed method saved 16 operations for each 8 8 DCT blocks, while achieved almost the same video quality. In order to further reduce the computation of DCT-to-IT transform, the integer form of S d can be easily derived using the similar method described in [7], and the multiplications to compute S d also can be replaced by just using add operations and shift operations just like doing in [8].…”

“…That is the de-quantized MPEG-2 DCT coefficients are first converted to pixels data through inverse DCT and are then transformed to IT coefficients through IT. This process is also called the pixel domain method [7] [8]. Obviously, this method cannot be adopted in the transform domain transcoder since that the incoming MPEG-2 video sequence is already decoded to the pixel data.…”

“…In order to implement the MPEG-2 to H.264 transcoder efficiently in the transform domain, we should try our best to speed up the process of DCT-to-IT transform. Jun Xin et al, [7] and Bo Shen [8] have derived two different transform kernel matrices for DCT-to-IT transform, and have showed their methods outperform the pixel domain method respectively. Although Jun Xin's method needs 64 less operations compared with Bo Shen's method, there still exits 19 nonzero elements in his transform kernel matrix, which cause DCT-to-IT transform to need 352 multiplications and 352 additions (for total of 704 operations).…”

Video transcoding from H.264/AVC to MPEG-2 with reduced computational complexity

“…This is quite different from the type of transform domain conversion needed in MPEG-2 to H.264/AVC transcoding, which was addressed in [20], mainly because of intraprediction. While intraprediction can be avoided in DCT to IT conversion, i.e., MPEG-2 to H.264/AVC transcoding, it is unavoidable in IT to DCT conversion because an H.264/AVC stream to be transcoded is not supposed to be constrained in its intraprediction modes.…”

An Efficient Algorithm for VC-1 to H.264 Video Transcoding in Progressive Compression

“…The techniques proposed in the literature attempt to reuse the information gathered during the decoding stage to improve the transcoder performance. The transform domain approaches convert MPEG-2 DCT coefficients to H.264 transform coefficients [4,5]. These techniques, however, have limited applicability due to the fact that H.264 uses directional predictors in intra coding and determining the optimal predictor in the transform domain is computationally expensive.…”