Title :
A practical software-only noncausal predictive video codec for low-bit-rate multimedia applications
Abstract :
A new coding scheme for bit rates below 150 Kb/s is proposed. The scheme combines (i) three-dimensional (3D) noncausal recursive prediction, (ii) vector quantization, and (iii) conditional replenishment. The 3D noncausal prediction model provides an alternative to causal (or unilateral) prediction, yet is not commonly used since it precludes recursive computations owing to its bilateral nature. This paper shows how to obtain a practical, near-optimal recursive implementation of the 3D noncausal video model and demonstrates its application in video compression. The proposed video codec is shown to produce high-quality compressed video at bit rates below 150 Kb/s. In contrast to the International Telecommunication Union (ITU) H.263 standard, the video codec exhibits acceptable video quality with a higher peak signal to reconstructed noise ratio (PSNR).
Keywords :
Application software; Bit rate; Code standards; ISO standards; PSNR; Predictive models; Telecommunication standards; Vector quantization; Video codecs; Video compression;
Journal_Title :
Electrical and Computer Engineering, Canadian Journal of
DOI :
10.1109/CJECE.2005.1541727