摘要
可独立解码的视频帧内编码帧(Ⅰ帧)具有作为随机访问点、防止错误传播等重要功能,因而在视频应用中扮演着至关重要的角色,现今一些高质量视频应用例如数字电影中也有直接使用全Ⅰ帧编码的情况。然而Ⅰ帧较低的压缩比与其重要程度形成了一对亟待解决的矛盾。近年来,最先进的帧内编码技术多是基于H.264/AVC国际视频编码标准所采用的混合视频帧内编码框架所提出的。因此,从理论上深入理解并完善信源编码理论在混合视频帧内编码框架中的应用,并在此基础上研究提出工程可实现的高效帧内编码算法具有十分重要的理论与现实意义。
在理论分析方面,本文认为混合视频帧内编码框架将一帧图像的信源熵用链式法则拆解为多个像素块条件熵之和,并通过帧内预测技术来去除像素块之间的互信息实现对条件熵的逼近；在编码每个像素块时,通过变换技术将块内像素映射为互不相关的变换域信源符号,最后应用熵编码技术对每个变换域信源符号逼近其信息熵。由于帧内预测技术和变换技术中所期望去除的信源符号间互信息很难定义和表达,因此工程上常常以线性相关性代替互信息作为技术优化目标,本文分析讨论了互信息与线性相关性的关系以及这种工程近似方法的合理性。
作为对上述相对抽象层面理论分析的延伸,本文更具体地研究了在一个特定的视频图像区域上如何获得最佳的单一模式帧内预测和变换的联合。事实上,最佳的帧内预测就是能完全去除参考像素与当前编码块目标像素之间线性相关性的维纳滤波器,而对应于该最佳预测的最佳联合变换即为在该预测所获得的残差上统计生成的KLT变换。
在理论研究的基础上,本文进一步针对帧内预测这一重要技术环节,提出了三种新的高效技术方案,这三种技术都应用了本文之前所讨论的在一个特定视频图像区域上生成最优帧内预测模式的算法。三种帧内预测技术方案有各自适合的应用条件,普通自适应帧内预测技术需要额外编码传输较大量的预测滤波器信息,因此主要适用于高分辨率且纹理细节复杂独特的视频序列；解码端导出自适应帧内预测模式技术无须传输额外滤波器信息,因而可适用于更宽广的分辨率范围,但对解码端的计算能力要求很高；固定预测滤波器系数技术的预测滤波器在线下统计训练得到因而具有最低的编解码复杂度,但其性能取决于编码序列和训练集之间的相似性,难以保证在所有类型的视频序列下都性能良好。实验证明,在对应的应用条件下,三种技术相比当前最先进的国际视频编码标准H.264/AVC均能取得明显的性能增益(尤其是在高分辨率纹理细节复杂的视频序列上)。
Independent decodability enables Intra coded frame to be used as random access point and preventer of error propagation, which further makes intra frame very essential to various video applications. However, the relatively low compression ratio of Intra coded frame and its importance formed an urgent conflict which needs to be solved. The recently proposed Intra coding technologies are mostly based on the hybrid intra coding framework adopted by H.264/AVC. Therefore, understand and perfect how the coding theory is applied to the hybrid intra coding framework and propose advanced intra coding algorithms based on the study of the theory is of substantial theoretical and practical significance
For theoretical analysis, this work perceives that the hybrid intra coding framework break down the entropy of a frame into the sum of the conditional entropy of many successive pixel blocks according to the chain rule. To reach the conditional entropy of the blocks, Intra Prediction is introduced to remove the mutual information of different blocks. When coding each block itself, Transform is applied to map the vector composed of pixels within the block to a vector composed of independent symbols in a transform domain. Finally, entropy coding is applied to each symbol in the transformed domain to reach its entropy independently. Since it's hard to really express and deal with mutual information, we usually take linear correlation as an approximation of mutual information to be used as the optimization goal. This work analyzed the relationship between mutual information and linear correlation as well as the correctness of the approximation.
As an extension of the theoretical analysis at the abstract level, this work studied how to generate the optimal single mode joint Intra Prediction and Transform for a specific region of image. In fact, the optimal Intra Prediction is the Wiener Filter that totally removes the linear correlation among reference pixels and the coded pixels, moreover, the jointly optimal Transform is the KLT for the prediction residues got from the optimal prediction.
Based on the theoretical analysis, this work further focused on Intra Prediction and proposed 3 types of advanced prediction schemes which all applied the algorithm of adaptively generating the optimal intra prediction mode for a specific image region. The 3 types of schemes are suitable for different application scenarios. The normal adaptive intra prediction scheme has to transmit extra information of the generated prediction filters therefore is mainly used for high resolution sequences with relatively complex texture. The decoder derived adaptive intra prediction scheme avoids transmitting the overhead thus can be used for a wider range of resolution, yet it introduces much calculation complexity on the decoder side. The fixed prediction filter scheme generates prediction filters by off-line statistics thus has lowest complexity both in encoder and decoder side, yet its performance is largely affected by the statistical similarity between training sequence set and the coded sequence. Experiments showed that under the suitable application scenarios, all 3 schemes achieved significant coding gain compared with H.264/AVC, especially for high resolution sequences with complex texture.
引文
[1]"Advanced video coding for generic audiovisual services," ITU-T Rec. H.264, Mar.2005.
[2]Taichiro Shiodera, Akiyuki Tanizawa, Takeshi Chujoh, "Bidirectional Intra Prediction"、VCEG 31st Meeting:Marrakech, MA、2007-1、 VCEG-AE-14、 page(s):1-6
[3]Shiodera, T., Tanizawa, A., Chujoh, T.,2007. Bidirectional Intra Prediction. ITU-T Q.6/SG16, C181, Geneva, Switzerland.
[4]Shiodera, T., Tanizawa, A., Chujoh, T.,2007. Improvement of Bidirectional Intra Prediction. ITU-T Q.6/SG16 VCEG, VCEG-AG08, Shenzhen, China.
[5]S.-L. Yu and C. Chrysafis、"New Intra Prediction using Intra-Macroblock Motion Compensation"、doc JVT-C151、2002-3
[6]Satoshi Kondo, Hisao Sasai, Shinya Kadono、"Tree structured hybrid intra prediction"、International Conference on Image Processing、2004-10、 Volume 1、 page(s):473-476
[7]Satoshi Kondo, "Video Coding with Hybrid Intra Prediction", Electronics and Communications in Japan (Part III:Fundamental Electronic Science), vol.90, Issue 2, Pages 50-62, February 2007
[8]T. K. Tan, C. S. Boon, and Y. Suzuki、"Intra Prediction by Template Matching" International Conference on Image Processing、2006-10、 Volume 1 page(s):1693-1696
[9]Jingheng Yang, Baocai Yin, Yanfeng Sun、"A Block-Matching Based Intra Frame Prediction for H.264/AVC"、International Conference on Mutimedia and Expo、 2006-7、Volume 1、page(s):705、708
[10]Johannes Balle, Mathias Wien. VCEG-AE11 《Extended Texture Prediction for H.264 Intra Coding》.31st Meeting:Marrakech, MA,15-16 January,2007
[11]Yi Guo. VCEG-AG17 《Priority-based Template Matching Intra Prediction)) 33rd Meeting:Shenzhen, China、20 October、2007
[12]T.K. Tan, C.S. Boon, and Y. Suzuki. 《Intra Prediction by Template Matching》 ICIP'06. Oct.2006 Page(s):1693-1696
[13]仇佩亮、“信息论与编码”、高等教育出版社、2003-12、第1版、6-59页
[14]Ahmed, N., Natarajan, T., Rao, K. R.,1974. Discrete Cosine Transform. IEEE Transactions on Computers, C-23(1):90-93.
[15]Jayant, N. S., Noll, P.,1984. Digital coding of waveforms. Prentice-Hall, Englewood Cliffs, New Jersey, p.62-64,524-546.
[16]Jizheng Xu, Feng Wu, Wenjun Zhang, "Intra-Predictive Transforms for Block-Based Image Coding", IEEE Transactions on signal processing, vol.57, No.8, August 2009, Page(s):2822-2825
[17]Yao Wang, J"orn Ostermann, Ya-Qin Zhang, "Video Processing and Communications", Prentice Hall,2001-9, chapter 9
[18]http://iphome.hhi.de/suehring/tml/download/KTA/
[19]G. Bjontegaard, "Calculation of Average PSNR Differences between RD curves", ITU-T SG16/Q6, Doc VCEG-M33, Austin, Texas, USA, April.2001
[20]I. Richardson, "H.264 and MPEG-4 Video Compression:Video Coding for Next Generation Multimedia," Wiley 2003.
[21]Wiegand T., Sullivan G.J., Bintegaard G.、"Overview of the H.264/AVC video coding standard"、Circuits and Systems for Video Technology、2003-7、Volume 13、Issue 7、page(s):560-576
[22]T. Shiodera, A. Tanizawa, and T. Chujoh, "Bidirectional Intra Prediction," ITU-T SG16/Q.6 VCEG, VCEG-AE14, Marrakech, Morocco, January 2007.
[23]Taichiro Shiodera, Akiyuki Tanizawa, Takeshi Chujoh, "Block Based Extra/Inter-Polating Prediction For Intra Coding," IEEE International Conference on Image Processing, vol.6, pp.445-448, Sep.2007
[24]Y. Ye and M. Karczewicz, "Improved intra coding," ITU-T Q.6/SG16 VCEG, VCEG-AG11, Shenzhen, China, October 2007
[25]Limin Liu, Yuxin (Zoe) Liu, and Edward J. Delp、"Enhanced Intra Prediction Using Context-Adaptive Linear Prediction"、Piture Coding Symposium、2007、 PCS v2007
[26]Peng Zhang, Debin Zhao, Siwei Ma, "Multiple Modes Intra-Prediction in Intra Coding"、International Conference on Multimedia and Expo、2004-6、Volume 1、 page(s):419-422
[27]Tomonobu Yoshino. VCEG-AG06《An extended intra prediction scheme based on frequency component representation》.33rd Meeting:Shenzhen, China.20 October,2007
[28]Matthias Narroschke、"Extending H.264/AVC by an adaptive coding of the prediction error“、Piture Coding Symposium、2006、25th PCS v2006、page:6
[29]Yumi Sohn, Woo Jin Han. 《One Dimensional Transform for H.264 Based Intra Coding》.PCS2007
[30]Xiaoyang Wu, Sijia Chen, Qichao Sun、"Derivation for Adaptive Scan of Intra Prediction in Video Coding"、Piture Coding Symposium、2007、PCS v2007
[31]Zhang Nan, Yin Baocai, Kong Dehui、"Spatial Prediction Based Intra-Coding"、 International Conference on Multimedia and Expo、2004-6、 Volume 1、 page(s):97-100
[32]Shohei Matsuo, Seishi Takamura, Kazuto Kamikura. VCEG-AF05 《Extension of Intra Prediction Using Multiple Reference Lines》.32nd Meeting:San Jose, CA, USA,20-21 April,2007
[33]Yung-Lyul Lee, Ki-Hun Han, Gary J.Sullivan、"Improved Lossless Intra Coding for H.264/MPEG-4 AVC”、IEEE Transactions on Image Processing,2006-9, Volume 15、 Issue 9、 page(s):2610-2615
[34]Armando J. Pinho and Antonio J.R.Neves. 《L-Infinity Progressive Image Compression). PCS2007》
[35]毕厚杰、“新一代视频压缩编码标准——H.264/AVC",人民邮电出版社、2005-5、第1版、1-6页,92-96页,210-219页
[36]Iain E. G. Richardson、 "Video Codec Design——Developing Image and Video Compression Systems"、 John Wiley & Sons, LTD、1st Edition, page(s):1-295
[37]Gormish, M.J. Lee, D. Marcellin、"JPEG2000:overview, architecture, and applications"、International Conference on Image Processing、2000-9、 Volume 2、 page(s):29-32