5. 結果與討論
5.2. Error Propagation Effects
This section presents the frame-by-frame comparison of error propagation effects using different MDC methods. The effects of error propagation were examined for a single frame loss occurring at different hierarchical levels of Mobile sequence at QP=28. We use one packet for each frame in each descriptor and the error propagation results of a single packet loss at different hierarchical levels are shown in Fig.7, where we renumber the selected frames according to decoding order.
Figs.7(a-c) show the results of the frame loss occurring at levels 0, 1 and 2, respectively. In Fig.7, y-axis denotes PSNR degradation and x-axis the frame number (in decoding order). From Fig.7(a) it is observed that almost all the methods suffer from severe error propagation for the P-frame loss, except the proposed (S+T+D). This is due to that the proposed (S+T+D) duplicates key-frames to
(a) Frame loss at level 0 (the 2nd frame is lost)
(b) Frame loss at level 1 (the 3rd frame is lost) (c) Frame loss at level 2 (the 4th frame is lost)
Fig.7 Frame-by-frame comparison.
10
two descriptors and thus, when only one of them is loss, the other can be used to reconstruct frame without quality degradation and error propagation. In both proposed (S) and proposed (S+T) methods, key-frames are spatially split to two descriptors and hence, the P-frame loss in one descriptor will cause partial-frame loss which is recovered by using spatial estimation, suffering from quality degradation and error propagation. As for default_QP, although it duplicates the entire sequence to two descriptors, the same frames in the two descriptors are at different levels and thus, the lost key frame can only be recovered by the corresponding low quality frame in the other descriptor. This also results in quality degradation and error propagation. It is worth to mention that even though the quality degradation of default_QP in Fig.7(a) is smoother than those of proposed (S) and proposed (S+T), it is at the cost of bit-rate redundancy. That is why default_QP has worse performance than proposed (S) and (S+T) as shown in Fig.6. Compared with default_QP, modified_QP suffers from much severe quality degradation because the top-level frame used to recover the lost key frame has been set to QP=51 to reduce the bit-rate. Compared with Fig.7(a), the results in Figs.7(b) and (c) show that when the fame loss occurs at level 1 or 2, the error propagation effects are substantially reduced for all the methods and the performance gaps between different methods are also decreased. To summarize, the results in Fig.7 show that quality degradation and error propagation in the hierarchical prediction structure are affected by key frames most, and level-1 and level-2 frames the second. By taking into account the importance of frames at different levels, proposed (S+T+D) optimizes the trade-off between coding efficiency and error resilience and achieves the overall best performance.
參考文獻
[1] O. Campana, R. Contiero, “An H.264/AVC Video Coder Based on Multiple Description Scalar Quantizer,” IEEE Asilomar Conference on Signals, Systems and Computers (ACSSC), 2006.
[2] R. Bemardini, M. Durigon, R. Rinaldo, L. Celetto, and A. Vitali, “Polyphase Spatial Subsampling Multiple Description Coding of Video Streams with H.264,” Proceedings of IEEE Intel. Conf. on Image Processing (ICIP), Oct. 2004.
[3] J. G. Apostolopoulos, “Error-Resilient Video Compression Through the Use of Multiple States,” Proceedings of IEEE Intel. Conf. on Image Processing (ICIP), Vol. 3, 2000.
[4] C. W. Hsiao and W. J. Tsai, “Hybrid Multiple Description Coding Based on H.264,” IEEE Trans. on Circuits and Syst. for Video Technol., Vol. 20, No.1, Jan. 2010.
[5] H. Schwarz, D. Marpe , T. Wiegand, “Analysis of hierarchical B pictures and MTCF,”
IEEE International Conference on Multimedia & Expo, ICME ’06, pp. 1929-1932
[6] C. Zhu and M. Liu, “Multiple description video coding based on hierarchical B pictures,”
IEEE Trans. Circuits and Systems for Video Technology, vol.19, No.4, April 2009.
11
[7] J. Reichel, H. Schwarz, and M. Wien, Joint Scalable Video Model 11 (JSVM 11), Joint Video Team, Doc. JVT-X202, Jul. 2007.
[8] H.264/AVC Ref. Software – JM, http://iphome.hhi.de/ suehring/tml/.
[9] W. J. Tsai and Hao-Yu You, “Multiple description video coding based on hierarchical B pictures using unequal redundancy,” To appear in IEEE Trans. on Circuits and Systems for Video Technology.
計畫成果自評
本計畫的目標是設計一架構於 Hierarchical B 上的多重描述演算法，並結合不對等保護 的觀念使整體 rate-distortion 效果能達最好。我們已依計畫目標完成方法的設計及相 關的實驗，而研究成果也已被國際期刊IEEE Trans. on Circuits and Systems for Video Technology 所接受[9]。
12
我們根據 hierarchical B 的架構，研究出不同於傳統的時間域錯誤隱藏方法，此錯誤隱 藏方法可分成兩部分:1.提供 RB-frame 的重建，2. 提供 NRB-frame 的重建，依此兩類 frame 的特性，所參照的 motion vector 種類有所不同；RB-frame 在錯誤隱藏中所需要 的 motion vector 是使用還未 decoded frame 的 motion vector 資訊，而 NRB-frame 則 是使用最鄰近 frame 的 motion vector 當作參照。
使用了上述所提到的 motion vector，我們在補償錯誤的方法中，以 pixel 為單位做雙 向 motion vector interpolation and extrapolation。實驗結果顯示，在發生 whole frame loss 的情況下，我們的方法能提供最佳的畫面品質。
完成各種錯誤隱藏方法在不同視訊內容特性與傳輸通道的實驗與分析
在實驗結果中，我們所提出的空間域錯誤隱藏方法對於不同類型的視訊內容，並沒有很 明顯的差別，residual 的錯誤補償的優劣主要是取決於視訊壓縮的 QP 大小，而不會因 為動態、靜態或複雜的視訊內容有明顯的變動。
時間域的錯誤隱藏方法則對於靜態的視訊內容能有較好的效益，相反的動態的視訊內容 因有太多大幅度的 motion vector，或是 scene change 的情況過多，都會直接影響到時 間域錯誤隱藏的效益，所以動態的視訊內容較不適用時間域的錯誤隱藏方法。
為了模擬不同的傳輸通道的情況，我們設計了 0%~20%的 random loss rate 來實驗，在 網路傳輸條件較為可靠時，整體效益的表現以時間域的錯誤隱藏方法為佳，但若是像無 線傳輸這類型會發生較高 loss rate，則以空間域的效益較高。
完成 hierarchical B 架構上，發生傳輸錯誤的實驗與分析
在 hierarchical B 的架構中，若是發生了傳輸錯誤，所造成的 error propagation 遠 比一般傳統架構中來的嚴重許多，在實驗結果中發現，不同層的錯誤發生都會有著不同 大小的影響程度，進而導致每層的擁有不同的重要性，皆需不同的保護機制，也就是 UEP(unequal error protection)，而這也是設計最佳切割演算法中考量的重要一環。
國科會補助計畫衍生研發成果推廣資料表
(英文) Multiple description coding method based on Hierarchical B-picture structure for video streaming applications 護的key frame level，把完整的key frame (I or P frame) 複製一份讓兩個描 述檔各自擁有一份以提供最完善的error resilience。「空間域切割」法是用於
(英文) Based on Hierarchical B-picture prediction structure, our multiple description coding (MDC) generates two descriptors by applying duplication for key frames (i.e., I or P-frame) which require the highest error resilience; apply spatial-splitting for Reference-B frames (RB-frame) which require modest error resilience; and apply temporal-splitting (T) for non-reference B- frames (NRB-frame) which require the lowest error resilience.
With the proposed MDC, different estimation methods are designed for different frames.
In case of one-descriptor loss in key frame level, the lost key frames can be found in the other descriptor; in case of one-descriptor loss in RB-frame level, the loss RB-frame can be estimated by using spatial estimation method; in case of one-descriptor loss in NRB-frame, the loss NRB-frame can be estimated by using temporal estimation method.
產業別
資訊服務業99 年度專題研究計畫研究成果彙整表
其他成果
(
無法以量化表達之成果如辦理學術活動、獲 得獎項、重要國際合 作、研究成果國際影響 力及其他協助產業技 術發展之具體效益事 項等，請以文字敘述填 列。)
本計畫的研究成果已為頂級國際期刊 IEEE Transactions on Circuits and Systems for Video Technology 所接受, 因此具國際影響力, 而所設計的方法 可實際用於視訊串流的應用中, 改善錯誤容錯能力, 因此具有產業的應用性.
成果項目 量化 名稱或內容性質簡述
測驗工具(含質性與量性) 0
課程/模組 0
電腦及網路系統或工具 0
教材 0
舉辦之活動/競賽 0
研討會/工作坊 0
電子報、網站 0
科 教 處 計 畫 加 填 項
目 計畫成果推廣之參與（閱聽）人數 0
國科會補助專題研究計畫成果報告自評表
請就研究內容與原計畫相符程度、達成預期目標情況、研究成果之學術或應用價 值（簡要敘述成果所代表之意義、價值、影響或進一步發展之可能性） 、是否適 合在學術期刊發表或申請專利、主要發現或其他有關價值等，作一綜合評估。
1. 請就研究內容與原計畫相符程度、達成預期目標情況作一綜合評估
■達成目標
□未達成目標（請說明，以 100 字為限）
□實驗失敗
□因故實驗中斷
□其他原因 說明：
2. 研究成果在學術期刊發表或申請專利等情形：
論文：■已發表 □未發表之文稿 □撰寫中 □無 專利：□已獲得 □申請中 ■無
技轉：□已技轉 □洽談中 ■無 其他：（以 100 字為限）
3. 請依學術成就、技術創新、社會影響等方面，評估研究成果之學術或應用價 值（簡要敘述成果所代表之意義、價值、影響或進一步發展之可能性）（以 500 字為限）
本研究成果可應用在實際的視訊串流應用中,因此有其應用價值,而所提的方法使用的技 術極為創新,已為國際期刊 IEEE Transactions on Circuits and Systems for Video Technology 所接受(尚未刊登), 因此極具學術價值.