Many modern video encoders use the Lagrangian rate-distortion optimization （RDO） algorithm for mode deci- sions during the compression procedure. For each encoding stage, this approach involves minimizing a cost, wh...Many modern video encoders use the Lagrangian rate-distortion optimization （RDO） algorithm for mode deci- sions during the compression procedure. For each encoding stage, this approach involves minimizing a cost, which is a function of rate, distortion and a multiplier called Lambda. This paper proposes to improve the RDO process by applying two modifications. The first modification is to increase the ac- curacy of rate estimation, which is achieved by computing a non-integer number of bits for arithmetic coding of the syntax elements. This leads to a more accurate cost computation and therefore a better mode decision. The second modification is to search and adjust the value of Lambda based on the char- acteristics of each coding stage. For the encoder used, this paper proposes to search multiple values of Lambda for the intra-4x4 mode decision. Moreover, a simple shift in Lambda value is proposed for motion estimation. Each of these modi- fications offers a certain gain in RDO performance, and, when all are combined, an average bit-rate saving of up to 7.0% can be achieved for the H.264/AVC codec while the same concept is applicable to the H.265/HEVC codec as well. The extra added complexity is contained to a certain level, and is also adjustable according to the processing resources available.展开更多
文摘Many modern video encoders use the Lagrangian rate-distortion optimization （RDO） algorithm for mode deci- sions during the compression procedure. For each encoding stage, this approach involves minimizing a cost, which is a function of rate, distortion and a multiplier called Lambda. This paper proposes to improve the RDO process by applying two modifications. The first modification is to increase the ac- curacy of rate estimation, which is achieved by computing a non-integer number of bits for arithmetic coding of the syntax elements. This leads to a more accurate cost computation and therefore a better mode decision. The second modification is to search and adjust the value of Lambda based on the char- acteristics of each coding stage. For the encoder used, this paper proposes to search multiple values of Lambda for the intra-4x4 mode decision. Moreover, a simple shift in Lambda value is proposed for motion estimation. Each of these modi- fications offers a certain gain in RDO performance, and, when all are combined, an average bit-rate saving of up to 7.0% can be achieved for the H.264/AVC codec while the same concept is applicable to the H.265/HEVC codec as well. The extra added complexity is contained to a certain level, and is also adjustable according to the processing resources available.
