臺灣博碩士論文加值系統

視訊精采片段的擷取以視訊處理的角度來看，是屬於高階事件(high level events)，這牽涉到一個人以主觀去認知一個事件，與傳統低階特徵(low level features)擷取有很大的差異。但在精采片段的擷取上，一些低階特徵依然是被需要的。首先我們必須對視訊內容的組成與一些特徵有足夠的瞭解，才有辦法完成高階的偵測與擷取。根據這些基本的特徵，才能夠對於高階事件有正確的理解。因此在此研究中所致力發展的目標有兩個：
一、鏡頭邊界偵測。
二、場景分類。
對處理分析視訊內容而言，自動化的鏡頭切割是很必要的一項工作。一般來說，鏡頭(shot)被視為組成視訊資料的基本單位，因此鏡頭邊界的偵測(shot boundary detection)是視訊內容分析的基礎，它可以提供我們做視訊摘要、檢索、瀏覽抑或是更高階視訊切割。鏡頭切割所作的工作中，最主要的就是將真正的鏡頭切換以及在同一個鏡頭內的正常變化區別出來。在完成鏡頭切換偵測之後，所要進行的工作就是針對這些分割好的鏡頭進行描述。我們的作法是先針對棒球影片定義出幾種經常出現的場景(scene)，再使用一些適當的低階特徵來幫助我們判斷出每一個切割好的鏡頭是屬於哪一類型的場景。每一種場景在球賽中各自有不同的意涵、出現的時機與內容。透過場景的類型我們可以對於球賽進行中的內容有更近一步的理解。相信有了場景為依據，在後續更高階的語意識別上可以有更精確的結果。

Highlight extraction belongs to high-level events in video processing, and the highlight is defined subjectively by human perception. For indexing video source, it is necessary narrow down the gap between high-level and low-level features. Therefore, we must understand the characteristics of content in video and its corresponding features in priori, In this paper. The research is focused on twofold:
1. Shot Boundary Detection
2. Scene Classification
In general, shots are regarded as a basic unit of video composition. Therefore, shot boundary detection is the basic of video analysis. Shot descriptor can be applied to many video applications such as video abstract, retrieval, index, browsing …etc. The main task is differentiated between real shot change and normal variation of a shot. After shot change detection, the following task is describing the segmented shot. In this paper, some scene types of game video is defined, then use proper low-level features to identify scene types of segmented shot. Each scene type includes different meaning, timing of appearance and content on game proceeding. With the help of scene types, people can understand further about the content of game proceeding. According to classified scene types, we can obtain more accurate result on follow-up high-level semantic recognition.

摘 要 I
Abstract III
誌謝 V
目錄 VI
圖目錄 X
表格目錄 XIV
公式目錄 XV
第1章 緒論 1
1.1 研究背景與動機 1
1.2 研究目標 3
1.3 論文結構 6
第2章 相關文獻回顧 7
2.1 鏡頭切換偵測技術 7
2.2 場景型態分類技術 10
第3章 鏡頭切換偵測 15
3.1 以模型為基礎的鏡頭切換偵測系統 15
3.1.1 鏡頭邊界的特性 15
3.1.2 特徵擷取 16
3.1.3 差異測量 18
3.1.4 系統流程圖 20
3.1.5鏡頭切換型態 20
3.1.6 各種鏡頭切換型態的視訊編輯模型 22
3.2 以模型為基礎的鏡頭切換偵測系統之探討 35
3.3 以模型為基礎的鏡頭切換偵測系統之修正 36
3.3.1 新模型定義 36
3.3.2 適應性門檻值(Adaptive Threshold) 40
3.3.3不可補償率(Un-compensated rate) 41
3.3.4 運動估測之快速演算法 43
3.4 新的鏡頭切換偵測架構 44
第4章 以樣本為基礎的棒球視訊場景型態分類 46
4.1 場景型態定義 46
4.2 特徵的探討與選用 48
4.2.1 場地地圖 49
4.2.2 使用學習向量量化(LVQ)進行場地分類 50
4.2.3 場地地圖之縮圖 55
4.2.4 色彩直方圖 58
4.2.5 場地地圖相似度計算 62
4.3 棒球視訊場景型態分類系統 63
第5章 實驗結果 67
5.1 程式介面 68
5.1.1 鏡頭切換偵測 68
5.1.2 棒球場景型態分類 70
5.2鏡頭切換偵測 72
5.2.1新模型驗證 72
5.2.2不可補償率驗證 76
5.2.3實驗數據 79
5.3 棒球場景型態分類 80
5.3.1土草分類訓練方式與內部測試準確度 80
5.3.2樣本資料庫的訓練 82
5.3.3實驗數據 84
第6章 結論與未來發展 88
參考文獻 90

[1]A. Nagasaka and Y. Tanaka, “Automatic video indexing and full-video search for object appearances,” Proceedings of the IFIP TC2/WG 2.6 Second Working Conference on Visual Database Systems II, pp. 113-127, September 1991.
[2]H. J. Zhang, A. Kankanhalli, and S. Smoliar. ”Automatic Partitioning of Full Motion Video,” ACM Multimedia Systems, Vol. 1, No. 1, pp. 10-28, 1993.
[3]S.W. Smoliar and H. J. Zhang, “Content based video indexing and retrieval,” IEEE Multimedia, Vol. 1, No. 2, pp. 62-72, Summer 1994.
[4]F. Arman, A. Hsu, and M. Y. Chiu, “Image processing on compressed data for large video databases,” Proceedings of the first ACM international conference on Multimedia, pp. 267-272, September 1993.
[5]B.-L. Yeo, and B. Liu, “Rapid scene analysis on compressed videos,” IEEE Transactions on Circuits Systems for Video Technology, Vol. 5, No. 6, pp. 533-544, December 1995.
[6]W.A.C. Fernando, C.N. Canagarajah, and D.R. Bull, “A unified approach to scene change detection in uncompressed and compressed video,” IEEE Transactions on Consumer Electronics, Vol. 46, No. 3, pp. 769-779, August 2000.
[7]S.M. Bhandarkar and A.A. Khombhadia, “Motion-based parsing of compressed video,” Proceedings of IEEE International Workshop on Multi-Media Database Management Systems, pp. 80-87, August 1998.
[8]L. Wang, B. Zeng, S. Lin, G. Xu, and H.Y. Shum, “Automatic extraction of semantic colors in sports video,” Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing, Vol. 3, pp. 617-620, May 2004.
[9]H. Lu and Y.P. Tan, “Unsupervised clustering of dominant scenes in sports video,” Pattern Recognition Letters, Vol. 24, pp. 2651-2662, November 2003.
[10]H. Lu and Y.P. Tan, “An unsupervised approach to dominant video scene clustering,” Proceedings of IEEE International Symposium on Circuits and Systems, Vol. 2, pp. 680-683, May 2003.
[11]S.C. Pei and F. Chen, “Semantic scenes detection and classification in sports video,” 16th IPPR Conference on Computer Vision, Graphics and Image Processing.
[12]H.H. Wang “A Robust Model-Based Technique for Video Scene Change Detection,” Master Thesis, 2003
[13]C.Y. Liao “Fast Correlation Block Matching Based on Partial Distortion Algorithm for Video Compression,” Master Thesis, 2003
[14]J. Luo, M. Boutell, R.T. Gray, and C. Brown “Image Transform Bootstrapping and Its Applications to Semantic Scene Classification,” IEEE Transactions on Systems, Man, and Cybernetics - Part B: Cybernetics, Vol. 35, No. 3, pp. 563-570, June 2005
[15]F. Naccari, S. Battiato, A. Bruna, A. Capra and A. Castorina “Natural Scenes Classification for Color Enhancement,” IEEE Transactions on Consumer Electronics, Vol. 51, No. 1, pp. 234-239, February 2005
[16]D. Xu, C. Bailey and R. Sotudeh, ”An improved three-step search block-matching algorithm for low bit-rate video coding applications,” Proceedings of IEEE International Symposium on Signals, Systems, and Electronics, pp. 178-181, October 1998.
[17]T. Kohonen, “The self-organization map,” Proceedings of the IEEE, Vol. 78, No. 9, pp. 1464-1480, September 1990