Abstract
The H.264/AVC video coding standard uses multiple reference frames and variable-size macroblock partitions in inter-prediction. This flexibility allows the standard to achieve excellent rate-distortion performance at the cost of high encoding complexity. We present an algorithm for fast joint selection of the reference frames and macroblock partitions in rate-distortion-based coding. Experimental results for eight standard test video sequences show that, compared with exhaustive search, our algorithm can save up to 74.9 % of the encoding time with negligible loss in rate-distortion performance. The presented algorithm is also evaluated using a subjective assessment metric; quality evaluation measures based on user satisfaction. Diverse types of video sequences were used with different frame rates, quantization parameters, and resolutions. The effects of changing bit rate and resolution on compression efficiency and viewers' satisfaction are also presented. Results show that our algorithm provides high scores of perceptual satisfaction that are significantly affected by the compression technique. As a result, we claim that our algorithm presents original and significant enhancement compared with exhaustive search. In addition, overall test results showed that our technique outperformed three of the best previously proposed methods and gave higher viewer satisfaction.