基于HEVC的码率控制算法

发布时间：2018-07-20 15:52

【摘要】：随着人们对视频服务消费水平的提高,人们也热衷于追求高分辨率视频。视频分辨率的增加,给视频的存储和传输带来了很大的挑战。目前,如何使编码产生的码流符合带宽要求是视频编码中的研究热点。码率控制成为了解决此问题的关键,码率控制通过控制编码量化参数,使输出码率接近带宽大小,充分利用带宽。本文通过研究新一代视频编码标准HEVC的帧内码率控制算法,提出了相关的改进算法和策略。主要包括两部分:1)HEVC帧内码率控制算法并没有考虑视频内容的复杂度,率失真性能还可以进一步提升。针对此问题,本文提出了基于复杂度分析的HEVC帧内码率控制算法。根据梯度因子来表征编码帧复杂度,梯度越大说明帧的复杂度越高,在分配时需要更多的比特来维持质量。首先通过图像复杂度和比特之间的近似线性关系,为每一帧合理分配比特。再根据帧内实际分配比特与预分配比特之差自适应更新平滑窗口,使帧与帧之间的比特波动减小。最后结合基本单元层的码率控制算法及时调节码率。与Li算法相比,对于相同的PSNR值,所提算法平均可以节省2.3%的码率,最大可以节省2.9%的码率。视频的客观质量平均可以提高0.08dB,最大可以提高0.12dB。整体码率波动性比Li要好,比特误差进一步减少。2)为了有效解决HEVC中的码率控制算法没有考虑视频特征保护的问题,提出了基于特征保护的HEVC码率控制算法。通过获取特征点掩膜,结合HEVC码率控制算法的三层比特分配框架。根据特征点分布情况给每个LCU分配不同的权重,提高LCU的匹配效率。在不显著降低视频主客观质量的前提下,尽可能提高视频特征点的保护数量。实验结果表明,所提算法与平台HM15.0相比,在时间复杂度可接受范围内,匹配率有所提高,平均提高1.27%,保护了更多的特征点,提升了算法的性能。
[Abstract]:With the increasing consumption of video services, people are also keen to pursue high-resolution video. The increase of video resolution brings great challenges to the storage and transmission of video. At present, how to make the code stream meet the bandwidth requirements is a hot topic in video coding. Rate control is the key to solve this problem. By controlling the coding quantization parameters, the output bit rate is close to the bandwidth and makes full use of the bandwidth. In this paper, a new video coding standard, HEVC, is studied, and some improved algorithms and strategies are proposed. It mainly includes two parts: 1) the rate control algorithm in HEVC frame does not consider the complexity of video content, and the rate-distortion performance can be further improved. To solve this problem, this paper proposes an intra-frame rate control algorithm for HEVC based on complexity analysis. The larger the gradient is, the higher the frame complexity is, and more bits are needed to maintain the quality in the allocation. Firstly, the image complexity and the approximate linear relationship between the bits are used to allocate the bits reasonably for each frame. Then the smoothing window is updated adaptively according to the difference between the actual allocation bits and the pre-allocated bits in the frame to reduce the bit fluctuation between frames. Finally, the rate is adjusted in time with the rate control algorithm of the basic unit layer. Compared with Li algorithm, the proposed algorithm can save an average rate of 2.3% and a maximum rate of 2.9% for the same PSNR value. The objective quality of video can be increased by 0.08dB on average, and the maximum by 0.12dB. In order to effectively solve the problem that the rate control algorithm in HEVC does not consider video feature protection, a rate control algorithm based on feature protection for HEVC is proposed. By obtaining feature point mask and combining with HEVC rate control algorithm, a three-layer bit allocation framework is proposed. According to the distribution of feature points, different weights are assigned to each LCU to improve the matching efficiency of LCU. The number of video feature points is improved as much as possible without significantly reducing the subjective and objective quality of video. The experimental results show that compared with HM15.0, the proposed algorithm improves the matching rate in the acceptable range of time complexity, increases the average value by 1.27, protects more feature points and improves the performance of the algorithm.
【学位授予单位】：宁波大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81

【参考文献】