Research On Decreasing The Computational Complexity Of HEVC Intra Coding

The the new video coding standard High-efficiency Video Coding （HEVC） achieves its goal, with the same perceptual video quality as H.264, HEVC can achieve a bit rate reduction of approximately 50%.To achieve a high compression rate, HEVC adopts more flexible coding structure and larger CU size than H.264 does. And so the encoder is expected to be increasingly complex. Coding unit partition and mode decision in HEVC introduces greater complexity which is difficult to meet the real-time need of coding applications. Therefore, this paper focuses on research on decreasing the computational complexity of HEVC Intra Coding.To reduce the computational complexity, a novel fast algorithm is proposed for CU partition in intra coding. A rough minimum depth prediction of the largest CU method and an early termination method for CU partition based on the total coding bits of the current CU are employed.（1） We found that the final CU depth is strongly correlated with the texture complexity of the input image in the HEVC intra encoder. So, our next work is to find the token parameter to discirbe the image complexity. On the basis of the statistics related to the different sizes of CUs with different QPs, we find the number of total coding bits that are used to encode a CU can make it well. We choose a group of thresholds for CU based on the total number of coding bits. For the CUs with same size and fixed QP, if the number of total coding bits of the CU are less than the selected threshold, then we terminate the CU partition early; otherwise, we let the partition continue. In the experiment, the proposed algorithm reduces encoding time by 32%on average and achieves an approximately 1.1% increase in Bj（?）ntegaard delta bit rate.（2） In the course of the experiment we found most sequences achieve good results, except for the sequence BlowingBubbles₄16x240, PartyScene₈32x480, BQSquare₄16x240. Because these sequences just have a few CUs to early terminate, that is, their depths are too big. To overcome this problem, we proposed the method of rough LCU minimum depth prediction to skip some certain CUs and thereby reduce the complexity of the encoder. We predict the minimum depth level of the current LCU compared with the top or left neighbor LCU. In the subsequent CU Splitting and pruning process we skip the RD cost checking of the CU with a depth level that is less than the predicted minimum depth level.Combined with the methods mentioned above, the proposed algorithm reduces encoding time by 45% on average and achieves an approximately 1.1% increase in Bj（?）ntegaard delta bit rate and a negligible peak signal-to-noise ratio loss.