Efficient transcoding of an MPEG-2 bit stream to an H.264 bit stream

Algorithm: The proposed research extracts the motion vectors and the transform coefficients from the incoming MPEG-2 bit stream. In the case of I frames, a standard deviation based fast mode decision process is applied to the transform coefficients. The pattern of the standard deviation of the 8x8 blocks within the each macroblock is used to determine the activity of the macroblock and then compute mode decisions compatible with the H.264 specification. In the case of P and B frames, the extracted motion vectors are refined over a one pixel window up to quarter pixel resolution and reused. This allows accommodating quarter pixel motion vectors as compared to half pixel motion vectors from MPEG-2 and it also allows reducing the effect of any errors due to the lossy quantization processes used. Inter frames also use a monotonic mode decision hierarchy, which does not force the MPEG-2 mode decisions yet reduces the computational complexity as compared to the full search motion estimation and mode decision process by 50%. (Abstract shortened by UMI.)...