Complexity reduction of motion estimation in HEVC.

摘要

The High Efficiency Video Coding (HEVC) standard is the latest video coding project developed by the Joint Collaborative Team on Video Coding (JCT-VC) which involves the International Telecommunication Unit (ITU-T) Video Coding Experts Group (VCEG) and the ISO/IEC Moving Picture Experts Group (MPEG) standardization organizations. The HEVC standard is an optimization of the previous standard H.264/AVC (Advanced Video coding) with the bit-rate reduction of about 50% at the same visual quality.;The most complex and time consuming process in the HEVC encoding is the motion estimation. The process involves finding the best matching block in the current frame by comparing it with a reference frame. Unlike, H.264 which had fixed sized blocks, HEVC has variable sized blocks which reduce the number of bits required by certain blocks in the frame where there is no motion change. But still the process of finding the best match is very time consuming and imposes computational complexity. Various algorithms like three-step search, diamond search and square search have been developed to reduce the computational complexity of the motion estimation module. The complexity can be further reduced by using an early termination technique to end the search process once it reaches a certain threshold. In this thesis, an algorithm is proposed for early termination of the search points by calculating a threshold. The algorithm is based on the predicted motion vector and the sum of absolute differences of the predicted motion vector for the search points. It is observed that if the prediction to the starting point is precise, then it can be used to calculate a threshold value and if any search point goes below the threshold, it can be declared as the best match. The experimental results based on the proposed algorithm tested on various video sequences show a reduction of the encoding time by about 5% to 17% with negligible Peak Signal to Noise Ratio (PSNR) loss ( less than 1 dB) as compared to the existing algorithm. The algorithm is more efficient for SD and HD resolution videos. The bit-rate increase is from 2% to 13.8 %. Metrics like Bjontegaard (BD)-PSNR and BD-Bit-rate are also used.