Directed acyclic graph based mode optimization for H.263 video encoding

Optimal mode selection for video coding is important in minimizing visual distortion given a rate constraint, and it has been studied in the literature using a single previous macroblock dependency assumption. We relax the assumption to multiple-macroblocks and develop a directed acyclic graph based procedure that solves the now generalized optimization problem. We demonstrate that in certain cases, multiple block dependencies can be accounted for while remaining computationally feasible. The optimization is sufficiently general that it can be applied to most multimode block-based video coding standards. Experiments show coding gain of up to 0.32 dB over University of British Columbia´s TMN10 implementation. We conjecture that the more general optimization can be more flexibly applied to irregular video objects, as encountered in multiple-VOP (video object plan) mode in MPEG-4