Font Size: a A A

Research On Rate-distortion Optimization For High Efficiency Video Coding

Posted on:2015-08-09Degree:DoctorType:Dissertation
Country:ChinaCandidate:S S WangFull Text:PDF
GTID:1108330479478716Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
With the constant infiltration of the internet technology for the daily life, the number and generation speed of the digital video increase rapidly, and big data times is coming. Massive video brought great challenge for the video storage and transmit, which makes the research on the video coding standard always be a hot topic in the academic and industrial field. In 2013, the new video coding standard, high efficiency video coding(HEVC) is formally published. Compared with the previous video coding standards H.264/AVC, the coding performance has been improved greatly.With the great performance improvement, the coding complexity of HEVC also increases greatly. Thus it makes much sense to optimize the video coding standard in order to further improve the coding efficiency in real application. Based on rate distortion theory, this paper mainly discusses the rate distortion optinmization for HEVC including rate control, intra coding, inter coding and perceptual coding optimization.First, the video transmission needs a good rate control algorithm to ensure the efficiency of the transmission. The present rate control algorithm in HEVC does not fully consider the new coding feature and new coding structure. In this paper, a novel rate control scheme is proposed considering the new coding structure and new features. Firstly, the rate distortion dependency based on Rate-GOP is investigated. Then rate model and distortion model for Rate-GOP are proposed based on the rate distortion dependency. Considering the mixed Laplacian distribution of the transformed coefficients, this paper proposed a transform domain based quadratic ρ-R model to establish the relation between R and QP. Based on the above models, an optimal rate allocation method is proposed. The experimental results show the proposed rate control algorithm has better rate control performance compared with the related rate control algorithms for HEVC.Second, more prediction modes are adopted in HEVC with the maximum number as 35 and quad-tree based partition scheme is utilized for the coding unit. They increase the intra coding complexity greatly. In order to reduce the intra coding complexity efficiently, based on the correlation among the gradient, video texture and prediction mode distribution, firstly, the paper proposed a shrinkage scheme for the number of the prediction mode; secondly, a prediction mode decision model is proposed to improve the prediction accuracy based on Hadamard transform and quantization. The experimental results show the proposed scheme can reduce the number of the prediction mode efficiently. And the intra coding complexity can be reduced greatly with negligible quality loss. Meanwhile, the proposed intra optimization scheme can also significantly reduce the intra coding complexity for AVS2.Third, for inter coding, HEVC adopts quad-tree based partition for the coding unit and multiple reference frame motion compensation to improve the inter coding efficiency, which increases the inter coding complexity greatly. In order to reduce the inter coding complexity efficiently, this paper proposed an optimized inter rate distortion scheme including the following two aspects: reference selection and early termination for the partition of coding unit. Firstly, based on the special structure of the reference frame set for HEVC, a fast reference frame selection scheme is proposed based on the motion complexity under the consideration of the reference frame distribution in order to reduce the complexity of the motion estimation. Secondly, based on the statistical results for the rate distortion cost for the coding unit with splitting and without splitting, a fast splitting scheme is proposed to determine whether to split the current coding unit. The experimental results demonstrate that the proposed schemes can reduce the inter coding complexity efficiently with negligible quality loss.Fourth, the influence factor for perceptual video quality of HEVC changes much due to the adoption of the new coding tools. This paper proposed a novel calculation scheme of perceptual factor based on the divisive normalization and a non-linear zoom scheme is incorporated to further determine the perceptual factor. Then the factor is used for the adjustment of the quantization parameter to meet the perceptual characteristic. Furthermore, a perceptual based rate distortion calculation scheme is proposed for the mode decision based on the perceptual characteristic. The experimental results show that the proposed method can resonablely adjust the quantization parameter and the subjective performance can be improved greatly. Meanwhile, the proposed perceptual based optimization scheme can also significantly improve the subjective video quality for AVS2.
Keywords/Search Tags:High efficiency video coding, rate distortion optimization, rate distortion dependency, rate control, gradient variance, motion complexity, perceptual characteristic
PDF Full Text Request
Related items