基于视觉感知的HEVC优化算法研究

发布时间：2018-04-19 18:25

本文选题：HEVC + 感知视频编码　；参考：《华侨大学》2017年硕士论文

【摘要】：近年来,随着多媒体信息和网络技术的高速发展,高清、超高清视频应用逐步走进人们的视野,给视频数据的存储和传输带来巨大挑战。在这种背景下,联合视频编码工作组(Joint Collaborative Team on Video Coding,JCT-VC)制定了新一代视频编码标准(High Efficiency Video Coding,HEVC)。HEVC与上一代国际视频编码标准H.264相比较,编码效率提高了一倍。但是和H.264一样,HEVC依然采用基于客观失真度量标准的拉格朗日率失真优化技术去优化其编码效率。然而,一方面其采用的失真度量方法误差平方和(Sum of Square Error,SSE)和平均绝对误差(Mean Absolute Differences,MAD)与人眼视觉相关度较低,另一方面拉格朗日算子仅仅与量化参数(Quantization Parameter,QP)相关,忽略了视频内容感知特性。也就是说,HEVC在优化其编码效率时未能充分考虑人眼视觉特性。由于人眼是视频最终接受者,在HEVC中考虑人眼视觉特性以提高其编码效率具有重要意义。为此,本文将围绕基于视觉感知的HEVC优化算法展开研究工作,具体如下:首先,由于率失真优化技术中拉格朗日算子忽略视频内容的感知特性,本文提出一种感知特征引导的率失真优化方法,以提高HEVC感知编码性能。考虑视频时空特性,本算法分别对每一编码树单元分别提取空域感知特征(梯度幅值比)和时域感知特征(梯度幅值相似度偏差比)。根据提取的感知特征自适应调整拉格朗日算子以引导率失真优化的整个过程过程。实验结果显示,和原始算法相比较,所提算法可很好地改善编码视频的感知质量。其次,提出一种基于感知的HEVC帧间码率控制算法。由于人眼视觉对视频图像的边缘信息比较敏感,该算法利用梯度幅值来有效描述图像边缘结构信息,引导单元层的比特分配,分配给人眼感兴趣区域更多码率。实验结果显示,所提算法与原始HEVC码率控制算法相比较,可以获得更好的感知性能。最后,提出一种自由能理论引导的HEVC帧内码率控制算法。根据自由能理论,利用局部方差度量各编码树单元的感知敏感度。然后根据各编码树单元的感知敏感度自适应的调整单元层比特分配方案。实验结果显示,与现有最新方法相比,本算法可以有效地提高HEVC感知编码效率。综上所述,本论文所提出的基于视觉感知的率失真优化算法,充分考虑人眼视觉特性和HEVC编码器特点,根据输入视频内容的感知特性优化码率资源分配,有效地提高了HEVC感知编码效率。本文所提算法对感知视频编码的发展与实际应用具有一定的理论依据和技术支撑。
[Abstract]:In recent years, with the rapid development of multimedia information and network technology, the application of high-definition and ultra-high-definition video has gradually come into people's vision, which brings great challenges to the storage and transmission of video data.Under this background, the Joint Collaborative Team on Video coding (JCT-VCC) has developed a new generation of video coding standard, High Efficiency Video coding (HEVCN). Compared with H.264, the coding efficiency of H.264 is twice as high as that of H.264.However, like H.264, HEVC still uses Lagrange rate-distortion optimization based on objective distortion criterion to optimize its coding efficiency.However, on the one hand, the error squared sum sum of Square error sum and mean absolute error mean Absolute difference (MADs) used by the method are lower than those of human eyes, and on the other hand, the Lagrange operator is only related to the quantization parameter Quantization (QP).The video content awareness feature is ignored.That is to say, HEVC does not fully consider human visual characteristics when optimizing its coding efficiency.Since the human eye is the ultimate receiver of video, it is very important to consider the human visual characteristics in HEVC to improve the coding efficiency.For this reason, this paper will focus on the HEVC optimization algorithm based on visual perception, as follows: firstly, because the Lagrangian operator ignores the perceptual characteristics of video content in rate-distortion optimization technology,In this paper, a rate-distortion optimization method based on perceptual feature guidance is proposed to improve the performance of HEVC perceptual coding.Considering the spatio-temporal characteristics of video, the algorithm extracts spatial perceptual features (gradient amplitude ratio) and time-domain perceptual features (gradient amplitude similarity deviation ratio) for each coding tree unit, respectively.According to the extracted perceptual features, the Lagrange operator is adaptively adjusted to guide the whole process of rate-distortion optimization.The experimental results show that compared with the original algorithm, the proposed algorithm can improve the perceptual quality of the coded video.Secondly, a perceptual rate control algorithm for HEVC interframe is proposed.Because human vision is sensitive to the edge information of video images, the gradient amplitude is used to effectively describe the edge structure information of the image, to guide the bit allocation of the unit layer, and to assign more bits to the region of interest to the human eye.Experimental results show that compared with the original HEVC rate control algorithm, the proposed algorithm can achieve better perceptual performance.Finally, a free energy guided HEVC intra-frame rate control algorithm is proposed.According to the free energy theory, local variance is used to measure the perceptual sensitivity of each coding tree unit.Then the cell layer bit allocation scheme is adaptively adjusted according to the perceptual sensitivity of each coding tree unit.Experimental results show that the proposed algorithm can effectively improve the efficiency of HEVC perceptual coding.To sum up, the proposed rate-distortion optimization algorithm based on visual perception in this paper fully considers the human visual characteristics and the characteristics of HEVC encoder, and optimizes bit rate resource allocation according to the perceptual characteristics of input video content.The efficiency of HEVC perceptual coding is improved effectively.The proposed algorithm has some theoretical basis and technical support for the development and practical application of perceptual video coding.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TN919.81

【相似文献】