This paper addresses the problem of bit-rate conversion of MPEG compressed video. We present selective requantization, a method for reducing the requantization errors in transcoding. The proposed method is based on avoiding critical ratios of the two cascaded quantizations (encoding versus transcoding) that either lead to larger transcoding errors or require a higher bit budget. Results show that selective requantization improve the quality of the transcoded images. The presented method is simple to implement and does not require side information.
Hani Sorial, William E. Lynch, André Vincen