臺灣博碩士論文加值系統

立體影像技術的成長一日千里，使得三維立體影像逐漸普及，與傳統二維影像最大的不同在於，三維影像需要至少兩個以上的影像序列以產生立體影像。雖然多個影像序列可提昇三維視訊的觀賞品質但也造成儲存資料量的增加，面臨資料傳輸、壓縮的負擔及困難。為了解決資料量龐大的問題，聯合視訊專家群(JVT)提出以鄰近視角間的相似性，提升壓縮效能降低整體位元率。然而，如何在壓縮位元率及影像品質間達到平衡，有效的位元率控制在三維視訊資料壓縮扮演著關鍵性的角色。
為了兼具影像品質及壓縮率進而提出有效的控制位元率方法，本論文提出以視覺感知的心理學理論分析三維視訊影像，發現人眼對於動態以及邊緣變化性高的物體有較高的興趣程度，藉此發展出以視覺感知為基礎的演算法。利用此觀念，將畫面中吸引人眼注視的物體藉由提高位元率提升該區域的畫面品質，反之容易忽略的區域藉由降低該區域的位元率降低，在不影響觀看品質的條件下有效節省位元率，達到在降低壓縮位元率及提昇觀賞影像品質的有效控制。
本論文提出之有效位元率控制方法是以視覺感知來定義人眼觀賞畫面之感興趣區域，其中利用目前編碼畫面的Macro-block(MB)與參考畫面的MB之間的平均亮度差值，透過just noticeable difference(JND)方程式定義具有高動量特性的興趣區域，另外以目前編碼畫面MB與參考畫面MB的亮度直方圖變化程度，定義該MB具有低動量的興趣區域及邊緣保留法描述外型明顯獨立於周圍的物件興趣區域。在正確描述興趣區域之後，根據不同的興趣程度，藉由調整量化參數(QP)提高興趣程度較高區域的位元率用以提昇畫面品質。
實驗結果顯示以平均亮度差值與亮度JND的比較、直方圖變化及邊緣保留法可以正確描述具有高動量、低動量及物件特性的興趣區域。實驗數據顯示本論文所提出的方法在相近的位元率的情況下感興趣程度較高的區域在PSNR上比JMVC提升約0.86dB的影像品質，在興趣程度較低的區域bit-rate節省2.5(bit/MB)。因此，本論文以視覺感知為基礎，有效的以三種方法描述畫面中的感興趣區域，在相近的位元率條件之下，本論文所提出的方法相較於JMVC在興趣區域有更好的畫面品質，並且在不感興趣區域以不影響觀看品質的條件下，節省了位元率，達到兼具影像品質及壓縮效率的目的位元率控制。

Currently, thanks for the development of video technology, the major topic of TV change from 2D to 3D. Compare to the traditional 2D video contents and 3D video contents, that need view content for 2D video and at least two or more than views for 3D video, respectively. Huge information of 3D video contents causes the awkward situation during data compression and data transmission. To solve the problem, Joint Video Team (JVT) proposes a coding method that utilizes the content-similar relationship between neighboring views to increase the efficiency of multi-view video coding (MVC). Therefore, rate control (RC) algorithm is the key technique to tradeoff data compression ratio and video quality of 3D video.
To control the bit-rate efficiently, considering video quality and compression ratio, the proposed algorithm utilizes the characteristic of visual perception theory from the analysis of 3D video that the region with high motion and individual object feature are interesting to viewer. Based on the concept, the proposed rate control algorithm gives the high bit-rate at the interesting region for preserve the image quality, decreases the image quality at the region of non-interesting for saving the bit-rate.
The proposed algorithm determines interesting region with high motion feature by using the relationship between luminance difference and Just Noticeable Distortion (JND) function at the current and reference macro-block (MB). To enhance the ability of describing motion feature, the proposed algorithm uses luminance histogram bin variation of the two MBs to describe low motion feature, edge strength (ES) describes the object with individual shape and obvious texture. By the interesting region determination, the rate-control algorithm is used by quantization parameter (QP) defining to control bit-rate at different levels of interesting region.
Simulation result of “PSNR of similar bit-rate” shows the proposed algorithm has better image quality than JMVC at the interesting region with high motion feature that increase 0.86dB of PSNR and save 2.5(bit/MB) at the non-interesting region.

目錄
第一章 緒論 1
1.1 簡介 1
1.2 研究動機與目的 1
1.3 論文架構 2
第二章 多視角編碼視訊系統 3
2.1 多視角視訊發展現況 3
2.2 多視角編碼架構 5
第三章　視覺感知理論 8
3.1 視覺感知概論 8
3.2 視覺感知特性 9
3.2.1 Searching for Patterns 9
3.2.2The Third Dimension 10
3.2.3Selectivity and Perceptual Constancy 11
3.2.4Environmental Factors 11
3.2.5Individual Differences, Purposes and Needs 11
3.2.6Context and Expectations 12
3.2.7Gestalt Principles of Visual Organization 12
3.2.7.1Similarity 13
3.2.7.2Continuation 13
3.2.7.3Closure 13
3.2.7.4Proximity 14
3.2.7.5Symmetry 14
3.2.8.The Moving Image 14
3.3 Just Noticeable Distortion 15
3.4 Ha’s Method 18
3.5 Wu’s Method 20
第四章 提出應用人類視覺感知於位元率控制之特性分析 21
4.1 前言 21
4.2 提出應用人類視覺感知應用於位元率控制之架構 21
4.3 提出並驗證視覺感知方法 25
4.3.1 Just Noticeable Difference (JND) 25
4.3.2 亮度直方圖 27
4.3.3 邊緣強度 30
4.4 整合視覺感知系統之方法 32
4.5 非對稱的應用 36
第五章 實驗結果 37
5.1前言 37
5.2 近似位元率PSNR比較 40
5.3 興趣區域的畫質比較 43
5.4 RD-curve 48
5.4.1 RD-curve 於主視角 48
5.4.2 RD-curve 於輔助視角 51
5.4.3 RD-curve 於全視角 58
5.5 Mean Opinion Scores(MOS) 61
5.6 SSIM比較 68
第六章 結論與未來展望 70
6.1 結論 70
6.2未來展望 70
第七章 參考文獻 71

[1]A.Vetro, Yea Sehoon, M. Zwicker, W. Matusik and H. Pfister, “Overview of Multiview Video Coding and Anti-Aliasing for 3D Displays,” IEEE International Conference on Image Processing, vol 1, pp. 17-20, 2007.
[2]H. Schwarz, D. Marpe, and T. Wiegand, “Overview of the Scalable Video Coding Extension of the H.264/AVC Standard,” IEEE Trans. Circuits Syst. Video Technology, vol. 17, no. 9, pp. 1103-1124, Sept. 2007.
[3]P. Merkle, A. Smolic, K. Muller, and T. Wiegand, “Comparative Study of MVC Prediction Structures,” ISO/IEC JTC1/SC29/WG11 and ITU-T SG16 Q.6, June 2007.
[4]C. Ha, W. Lee, S. Jin, and J. Jeong, “Human Perception of Asymmetrical 3-D Inputs,” 3DTV Conference, pp. 1-4, 2007
[5]G-L Wu, T-H Wu, Y-J Fu, S-Y Chien, “Perpcetual-Aware H.264/AVC Encoder with Hardware Perception Analysis, Conference on Multimedia and Expo (ICME), IEEE International , pp.790-795, 2010
[6]Y. L. Chang. “Algorithm and Architecture Analysis of the Video Signal Conversion for 2D to 3D Video,” Ph.D. dissertation, 2007.
[7]Polarized 3D vision method, IZ3D: http://www.iz3d.com/t-3dproductex.aspx
[8]I. Ahmad, “Multi-view Video : Get ready for Next-Generation Television,” IEEE Distributed System, vol. 8, Issue 3, pp. 1-6, Mar. 2007.
[9]Ying Chen, Ye-Kui Wang, Kemal Ugur, MiskaM. Hannuksela, Jani Lainema, and Moncef Gabbouj, “ The Emerging MVC Standard for 3D Video Services,”EURASIP Journal on Advances in Signal Processing, Vol. 2009, No. 1, January 2009.
[10]M. Tanimoto “Overview of Free Viewpoint Television,” Proc. IEEE Conference on Image Processing, vol. 21, no. 6, pp. 454-461, Jul. 2006.
[11]Toshiaki Fujii, Kensaku Mori, Kazuya Takeda, Kenji Mase, Masayuki Tanimoto, Yasuhito Suenaga, “Multipoint Measuring System for Video and Sound- 100-camera and Microphone System,” IEEE International Conference on Multimedia and Expo, pp. 437 – 440, July.2006.
[12]T. Fujii and M. Tanimoto, “Free-Viewpoint TV System Based on Ray-Space Representation,” SPIE ITCom , vol. 4864, pp. 175-189, Aug. 2002.
[13]Vicki Bruce, Patrick R. Green and Mark A. Georgeson, “Visual perception: physiology, psychology, & ecology” , Psychology Press,483pages, 2003
[14]Chandler Daniel, “ Notes on the Gaze”, 1995
[15]R. A. Kinchla and J. M. Wolfe, “The order of visual processing: “Top-down,” “bottom-up,” or “middle-out””, Attention, Perception, & Psychophysics , Volume 25, Number 3, 225-231
[16]Trevor MacInnis: http://upload.wikimedia.org/wikipedia/commons/thumb/4/47/B-2_Spirit_original.jpg/800px-B-2_Spirit_original.jpg
[17]Koffka K, “Principles of Gestalt Psychology,” New York, Harcourt Brace & Co.
[18]The Gestalt Principles: http://graphicdesign.spokanefalls.edu/tutorials/process/gestaltprinciples/gestaltprinc.htm
[19]Kalva H, Christodoulou L, Furht B, “Evaluation of 3DTV service using asymmetric view coding based on MPEG-2”, 3DTV Conference , pp.1-4, 2007
[20]C-H Chou and C-W Chen, “A perceptually optimized 3-D subband codec for video communication over wireless channels,” Transactions on Circuits and Systems for Video Technology, IEEE, vol. 6, Issue 2, pp.143-156, 1996
[21]C-H Chou and C-W Chen, “A perceptually tuned subband image coder based on the measure of just-noticeable-distortion profile,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 5, no. 6, pp.467-476, 1995
[22]Y. Liu, Z. G. Li, and Y. C. Soh. “A novel rate control scheme for low delay video communication of H.264/AVC standard,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 17, no. 1, pp. 66-78, Jan. 2007
[23]L. Narens , “ A note on Weber's law for conjoint structures,”Journal of Mathematical Psychology vol. 21, pp. 88–91, 1980
[24]3D Video Coding Technology ISO/IEC JTC1/SC29/WG11, MPEG2011/N12036, Geneva, Switzerland, March 2011.
[25]Z. Wang, A. C. Bovik, H. R. Sheikh, and E. P. Simoncelli , “Image quality assessment: From error visibility to structural similarity” IEEE Transaction on Image Processing, vol. 13, no. 6, pp.600-612, 2004.