Dirac was started off by British Broadcasting Corp. (BBC) in 2003 as an experimental video coding system based on wavelet technology, which is different from that used in the main proprietary/standard video compression systems. Over the years, Dirac has grown out of its initial development and it is now on offer as an advanced royalty-free video coding system designed for a wide range of users, from delivering low-resolution web content to broadcasting high-definition (HD) and beyond, to near-lossless studio editing. The Dirac’s video coding architecture and algorithms are designed with the “keep it simple” mindset. In spite of that the Dirac seems to give a two-fold reduction in bitrate over MPEG-2 for HD video and broadly competitive with state-of-the-art video codecs. This chapter introduces the architecture of Dirac video encoder. The overall encoding structure is discussed followed by the detail description of motion estimation, Overlapped Block-based Motion Compensation (OBMC), Discrete Wavelet Transform (DWT), Rate Distortion Optimization (RDO) quantization and entropy coding. The Dirac’s bitstream syntax for compressed video data storage and streaming is described. Besides that, the coding performance of Dirac in terms of compression ratio, PSNR, SSIM and VQM in comparison with H.264 as a reference are discussed. Related issues such as transcoding and streaming over packat erasure channel are also discussed.
Decoder-driven mode decision in a block-based distributed video codec
