H.264 encoder complexity is remarkable due to mainly variable block size Motion Estimation (ME) and exhaustive RDO mode decision.In this paper
statistical characteristics of modes among adjacent macroblocks spatially and temporally are first studied using H.264 reference software.Then a fast algorithm is proposed accordingly to reduce the computational load of the mode decision.Candidate modes for current macroblock are first inferred from given coded adjacent macroblocks by adopting motion information and ratios of defined mode sub-sets as inference measures.Then a partial RDO approach is applied to the candidate modes to make target mode selection.The ME is considered as a part of mode decision and no ME is required for particular modes that are eliminated in advance.The exhaustive RDO is invoked only if average distortion of the adjacent macroblocks goes beyond a predefined threshold.Simulation results demonstrate that
for image sequences of both class A and class B
the proposed algorithm can reduce encoder complexity significantly while coding quality is only slightly decreased.