An Efficient Mode Selection Prior to the Actual Encoding for H.264/AVC Encoder

Many video compression algorithms require decisionsto be made to select between different coding modes. In the case ofH.264, this includes decisions about whether or not motion compensationis used, and the block size to be used for motion compensation.It has been proposed that constrained optimization techniques,such as the method of Lagrange multipliers, can be used totrade off between the quality of the compressed video and the bitrate generated. In this paper, we show that in many cases of practicalinterest, very similar results can be achieved with much simpleroptimizations. Mode selection by simply minimizing the distortionwith motion vectors and header information produces verysimilar performance to the full constrained optimization, while itreduces the mode selection and over all encoding time by 31% and12%, respectively. The proposed approach can be applied togetherwith fast motion search algorithms and the mode filtering algorithmsfor further speed up.