Performance analysis of Dirac video codec in different motion vector accuracies and wavelet lifting decompositions

Abstract

Dirac is presented by BBC to compress video streams efficiently and eliminate software payments for license fees. Dirac video codec structure allows new wavelet tools to be specified by a user in order to improve its performance. In this paper, Dirac performance is analyzed for different motion vector accuracies and with different wavelet tools. Meanwhile, a new decomposition filter scheme is proposed for Dirac to compensate computational complexity that arises due to subpixel precision analyses and filter structures of this codec. Investigated codec has a stable structure therefore correct interpretations are possible for different test sequences using the comparison tables in this paper.