高效视频编码（HEVC）中快速帧内预测模式判断算法研究与实现

发布时间：2018-04-25 02:33

本文选题：HEVC + 帧内预测　；参考：《天津大学》2014年硕士论文

【摘要】：随着视频采集设备的广泛普及、人们对视频需求的不断上升,视频已渗透到人们生活的方方面面。然而有限的硬盘空间和网络带宽极大地制约了人们对视频文件的迫切需求,视频压缩成为必不可少的操作。为满足人们对高清、超高清视频的需求,新一代视频编码标准HEVC应运而生。与H.264/AVC相比,在相同的视频质量的条件下,HEVC可以节省50%的比特率。本文着重从编码结构、帧内预测、帧间预测三方面对HEVC的改进做了详细阐述。考虑到HEVC中33种角度预测模式分别代表了不同的预测方向,本文提出了一种基于DCT变换的快速帧内预测算法。该算法先对原始数据块进行DCT变换,利用DCT变换的性质,快速估计出待预测数据块的大致纹理方向,只把与该纹理方向距离较近的预测方向所对应的角度预测模式列为最优帧内预测模式的候选项,从而大大缩小了最优帧内预测模式搜索范围,达到加速帧内预测模块的目的。针对光场相机这一新兴的热点应用,本文从光场图像的自身特点出发,提出采用视频编码的方法压缩光场数据,预处理工作只需将光场图像转化为视频序列。本文探索了四种视频序列合成方法,并通过实验验证了基于微图像的视频序列合成方法的优越性。本文在ffmpeg开源库的基础上,实现了一套基于视频编码的光场数据编码器。该编码器支持用户指定视频序列合成方法、所用编码器类型以及编码参数等。
[Abstract]:With the wide spread of video capture equipment and the increasing demand for video, video has penetrated into every aspect of people's life. However, limited hard disk space and network bandwidth greatly restrict the urgent need for video files, and video compression becomes an essential operation. In order to meet the demand of HD and UHD video, the new generation video coding standard HEVC came into being. Compared with H.264/AVC, it can save 50% bit rate under the same video quality. In this paper, the improvement of HEVC is described in detail from three aspects: coding structure, intra-frame prediction and inter-frame prediction. Considering that 33 angle prediction modes in HEVC represent different prediction directions, a fast intra prediction algorithm based on DCT transform is proposed in this paper. In this algorithm, the original data block is transformed by DCT, and the approximate texture direction of the data block is estimated quickly by using the properties of DCT transform. Only the angle prediction mode corresponding to the prediction direction which is close to the texture direction is considered as the candidate of the optimal intra prediction mode, which greatly reduces the searching range of the optimal intra prediction mode and accelerates the intra prediction module. Aiming at the new hot application of optical field camera, this paper proposes a video coding method to compress the optical field data from the characteristics of the light field image. The preprocessing only needs to transform the light field image into a video sequence. In this paper, four kinds of video sequence synthesis methods are explored, and the superiority of the video sequence synthesis method based on microimage is verified by experiments. On the basis of ffmpeg open source library, a set of optical field data encoder based on video coding is implemented in this paper. The encoder supports user-specified video sequence synthesis methods, encoder types and encoding parameters.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】