面向高性能视频编码的变换编码技术研究

发布时间：2018-02-23 05:10

  本文关键词： 高性能视频编码 变换编码 自适应多变换 模板匹配 正交匹配追踪变换　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：20世纪以来,随着计算机图形图像技术,网络传输技术,多媒体技术的发展,基于视频技术的远程医疗、远程教育、视频会议等应用也越来越普遍。视频技术正在向大数据量,高分辨率的方向发展。视频大数据的传输和存储对视频技术的发展提出了挑战。因而高效率的视频压缩技术对视频技术的进一步发展显得尤为重要。2003年5月由运动图像专家组(MPEG)和视频编码专家组(VCEG)组成的联合编码小组JVT发布了当前主流的编码标准H.264/AVC。相比于之前的编码标准,H.264/AVC的视频压缩性能提升了 50%。十年之后,2013年1月由MPEG和VCEG组成的联合编码小组JCT-VC发布了新一代视频编码标准-高性能视频编码(HEVC)。HEVC在H.264的基础上性能又提升了 50%。现有的视频编码标准都是通过预测编码去除时间和空间的冗余得到残差信号,残差信号通过变换量化后进行熵编码。而残差的数据量占据整个编码数据量的绝大部分。因此一个好的变换编码方式将会大大提高编码器的压缩效率。无论是H.264还是HEVC,其变换都是采用固定的变换基,然而不同图像的纹理特性不一样,其预测残差的统计特性存在差异。对于不同统计特性的残差采取相同的变换方式,显然无法有效地消除残差的相关性。本文从变换编码出发,做了两方面的研究。一方面通过增加未来视频编解码器(Future Video Codec,FVC1)变换基的种类来改进FVC的变换编码效率。另一方面希望通过自适应的变换基来改进变换编码的效率。具体来说,本文的主要创新点如下:1、提出自适应多变换技术,在FVC现有的编码框架上,通过增加DCT-Ⅴ,DCT-Ⅷ,DST-Ⅰ三种变换基,编码器通过遍历不同的变换基找出一个最适合编码的变换基作为当前的变换基,自适应多变换技术为变换内容提供了不同的变换方式,相比于之前只有固定变换基的变换方式,自适应性更高。本文同时统计和分析变换基的使用情况,提出了快速的自适应多变换技术。实验表明,本文提出的自适应多变换算法可以提升压缩效率2%～3%左右。2、本文提出正交匹配追踪变换技术,在HEVC现有编码框架上,通过模板匹配,在线训练的方法得到与当前待变换内容相关的自适应变换基。然后通过与现有变换编码对比,找出一种最优的变换方式进行变换编码。相比于之前变换编码对于所有变换内容均采用相同的变换方式,正交匹配追踪变换技术考虑了变换内容的多样性。实验表明,本文提出的正交匹配追踪变换技术可以提升压缩效率1%～3%左右。
[Abstract]:Since 20th century, with the development of computer graphics and image technology, network transmission technology and multimedia technology, the application of tele-medicine, distance education and video conference based on video technology is becoming more and more common. The transmission and storage of video big data pose a challenge to the development of video technology. Therefore, the efficient video compression technology is particularly important for the further development of video technology. JVT, a joint coding group made up of moving image experts (MPEGG) and video coding expert group (VCE), released the current mainstream coding standard H.264 / AVC.Compared with the previous coding standard H.264 / AVC, the video compression performance of H.264 / AVC improved by 50. 10 years, January 2013. JCT-VC, a joint coding group composed of MPEG and VCEG, has released a new generation of video coding standards-high performance video coding / HEVCU. HEVC has improved the performance of H.264 based on H.264. existing video coding standards are all based on predictive coding removal time. And the redundancy of the space to get the residual signal, The residual signal is encoded by entropy after transform quantization, and the residual data amount accounts for the majority of the total encoded data. Therefore, a good transform coding method will greatly improve the compression efficiency of the encoder, regardless of H. 264. Or HEVCwhose transformation is based on a fixed transformation basis, However, the texture characteristics of different images are different, and the statistical characteristics of the prediction residuals are different. It is obvious that the correlation of the residuals can not be effectively eliminated by using the same transformation method for the residuals of different statistical characteristics. On the one hand, we improve the transform coding efficiency of FVC by adding the kinds of the future Video Codecor FVC 1) transform basis, and on the other hand, we hope to improve the efficiency of transform coding by adaptive transform basis. The main innovations of this paper are as follows: 1. Adaptive multi-transform technology is proposed. In the existing coding framework of FVC, DCT- 鈪，

本文编号：1526200