An efficient FGS to MPEG-1/2/4 single layer transcoder with R-D optimized multi-layer streaming technique for video quality improvement

Abstract

The need for interoperability poses a challenge with the widespread use of multiple video formats. A Multi Layer Streaming-Simplified DCT Domain Transcoder (MLS-SDDT) is proposed to achieve efficient transcoding and video quality improvement. However, an efficient architecture is needed to handle drifting errors. The MLS-SDDT algorithm addresses these issues by proposing: (1) an FGS compatible Simplified DCT Domain Transcoder (FGS-SDDT) architecture for MPEG-1/2/4 single layer transcoding, and (2) an R-D optimized multi-layer streaming technique for video quality improvement. To support efficient transcoding from FGS to MPEG-1/2/4, the FGS-SDDT adopts low complexity Simplified DCT Domain Transcoder (SDDT) architecture as an FGS compatible SDDT. To resolve the serious error-drifting problem from SDDT, especially for heterogeneous transcoding, a mathematical analysis of the error-drifting problem is provided and a solution by multi-layer streaming is adopted. The multi-layer streaming approach pre-computes the errors according to the mathematical analysis and encodes the errors as the second enhancement layer for error compensation. For different network conditions, an R-D optimized model is used to improve the bit allocation in the two enhancement layers for better R-D performance. Our experiments show that the MLS-SDDT can deliver 1.4 similar to 7.0 dB PSNR improvement for MPEG-1 and 1.9 similar to 8.6 dB improvement for MPEG-2 as compared to the SDDT. For FGS to MPEG-4 single layer transcoding, the MLS-SDDT can achieve minor PSNR loss (0.0 similar to 0.4 dB) with lower computational complexity as compared to the SDDT approach.