多媒體通信傳播品質藉由高效率視訊編碼(HEVC)技術顯著提升效能。為達到優良的編碼效率，HEVC使用了許多新的編碼技術，如:編碼單位(Coding Unit(CU))、預測單位(Predict Unit(PU))和轉換單位(Transform Unit(TU))，因此增大了運算複雜度。HEVC需要大量運算時間對所有的CU與PU區塊分割模式估算編碼效能以決定最好的區塊分割方式。針對此一高運算複雜度的問題，本論文提出快速CU與PU編碼模式決策的方法，在不降低品質的前提下減少HEVC編碼運算複雜度。在我們的方法中，基於CU切割流程中率-失真衡量數據(RD-Cost)、鄰居區塊的深度資訊，以及當前區塊資訊，提出了兩種提早終止(early termination)CTU最佳結構決策流程的方法。其中方法一基於「RD-Costs」、「鄰近區塊深度資訊」，並使用類神經網路將所得前述所得之資訊作為輸入特徵預測當前深度CU是否適合提前終止；方法二基於「當前區塊PU 2N×2N之資訊」與「前一層深度之切割情況」作為類神經網路之輸入，利用此類神經網路來決策當前PU是否需要測試更小之深度，最後將此兩種方法結合成一快速HEVC決策方法。實驗結果顯示，我們所提出的演算法，整體而言在BDBR僅上升1.87%的情況下，編碼時間上可以達到64.41%的加速效果，達到大幅度降低HEVC運算複雜度之目的。

The high efficienct video coding standard, HEVC, was proposed to enable high quality multimedia commnications. To achieve high coding efficiency, it has to determine the best coding unit (CU), prediction unit (PU), and transform unit (TU) through exhaustive search. It is time consuming and requires to speedup the CU and PU mode decisione process without degrading the coding quality. In this thesis, we proposed to speed up the HEVC inter-frame CU and PU mode decision process, in which rate-distortion cost, coding depth level of neighboring blocks, and current block information, are adopted as the input parameters of machine learning function to predict whether current depth CU an PU need to split or not. To fast decide the coding mode, it can: (1) utilize depth information and rate-distortion costs of both the neighboring CU and the current PU as the input parameters of neural network to predict whether the current depth CU need to split or not; (2) utilize RD-cost of the current block PU 2N2N and MSM mode, the upper depth information, such as whether the upper depth PU split or not, and the upper depth best mode RD-costs, as the input of neural network to determine whether the current depth PU mode needs further partition or not. In this research, both fast mode decision strategies are combined to yield a fast HEVC mode decision method. Experiments showed that the proposed method can reduced the encoding time on the average with 64.41%, and increases BD-bitrate about only 1.87%, as compared to the standard HEVC codec, HM13.0.

[1] G. J. Sullivan, et al, “Overview of the High Efficiency Video Coding (HEVC) Standard,” IEEE Trans. Circuits and Systems for Video Technology, vol. 22, no. 12, pp. 1649-1668, Sep 2012.
[2] T. Wiegand et al, “Overview of the H.264/AVC video coding standard,” IEEE Trans. Circuits Sys. Video Technology, vol. 13, no. 7, pp. 560-576, July 2003.
[3] J. Vanne et al, “Efficient mode decision schemes for HEVC inter prediction,” IEEE Trans. Circuits Sys. Video Technology, vol. 24, no. 9, pp. 1579-1593, Sept. 2014.
[4] HEVC Test Model.
[5] http://www.theregister.co.uk/2013/04/11/feature_wtf_is_h265_hevc/
[6] K. Choi and Euee S. Jang “Fast coding unit decision method based on coding tree pruning for high efficiency video coding,” SPIE Optical Engineering, vol. 51,Issue 3, March 2012.
[7] J. Lainema, et al “Intra Coding of the HEVC Standard,” IEEE Trans. Circuits Sys. Video Technology, vol. 22, no. 12, pp. 1792-1801, Dec 2012.
[8] http://www.hindawi.com/journals/ijrc/2012/473725/fig1/
[9] X.-L. Tang, S.-K. Dai, and C.-H. Cai, “An analysis of TZ search algorithm in JMVC,” IEEE Conf. Green Circuits and Systems (ICGCS), pp. 516-520, 2010.
[10] J.-F. Hu et al, “Speeding up the decisions of quad-tree structures and coding modes for HEVC coding units,” Institute of Computer and Communication Engineering Department of Electrical Engineering NCKU, June 2012.
[11] http://blog.sina.com.cn/s/blog_520811730101m9y2.html
[12] C. Cortes, and V. Vapnik. “Support-vector networks,” Machine learning, vol. 20, no. 3, pp. 273-297, 1995.
[13] LIBSVM -- A Library for Support Vector Machines. https://www.csie.ntu.edu.tw/~cjlin/libsvm/
[14] C. Seunghyun and K. Munchurl, “Fast CU splitting and pruning forsuboptimal CU partitioning in HEVC intra coding,” IEEE Trans. Circuits Sys. Video Technol., vol. 23, no. 9, pp. 1555–1564, Sep. 2013.
[15] L. Jong-Hyeok, P. Chan-Seob, and K. Byung-Gyu, “Fast coding algo-rithm based on adaptive coding depth range selection for HEVC,” in Proc. IEEE Int. Conf. Cons. Electron.-Berlin, Sep. 2012, pp. 31–33.
[16] K. Goswami, B. G. Kim, D. Jun, S. H. Jung, and J. S. Choi, “Earlycoding unit (CU) splitting termination algorithm for high efficiencyvideo coding (HEVC),” Electron. Telecommun. Res. Inst. J., vol. 36, no. 3, pp. 407–417, 2014.
[17] G. Correa, P. Assuncao, L. Agostini, and L. A. da Silva Cruz,“Complexity control of high efficiency video encoders for power-constrained devices,” IEEE Trans. Consum. Electron., vol. 57, no. 4,pp. 1866–1874, Nov. 2011.
[18] J. Vanne, M. Viitanen, and T. D. Hamalainen, “Efficient mode decisionschemes for HEVC inter prediction,” IEEE Trans. Circuits Sys. Video Technol., vol. 24, no. 9, pp. 1579–1593, Sep. 2014.
[19] T. Zhao, Z. Wang, and S. Kwong, “Flexible mode selection andcomplexity allocation in high efficiency video coding,” IEEE J. Topics Signal Process., vol. 7, no. 6, pp. 1135–1144, Dec. 2013.
[20] T. Guifen and S. Goto, “Content adaptive prediction unit size decisionalgorithm for HEVC intra coding,” in Proc. Picture Coding Symp. May 2012, pp. 405–408.
[21] H. Wei-Jhe and H. Hsueh-Ming, “Fast coding unit decision algorithm forHEVC,” in Proc. Asia-Pacific Signal Inf. Process. Assoc. Annu. Summit Conf. Oct. Nov. 2013, pp. 1–5.
[22] S. Liquan, Z. Liu, X. Zhang, W. Zhao, and Z. Zhang, “An effective CU size decision method for HEVC encoders,” IEEE Trans. Multimedia, vol. 15, no. 2, pp. 465–470, Feb. 2013.
[23] R. Garcia et al, “HEVCdecision optimization for low bandwidth in video conferencing applica-tions in mobile environments,” in Proc. IEEE Int. C. Multimedia Expo Workshops, Jul. 2013, pp. 1–6.
[24] X. Shen and L. Yu, “CU splitting early termination based on weighted SVM,” EURASIP J. Image Video Process., vol. 2013, no. 4, pp. 1–11, 2013.
[25] G. Bjontegaard, “Calcuation of average PSNR differences between RD-curves,” Doc. VCEG-M33 ITU-T Q6/16, Austin, TX, USA, 2-4 April 2001.