追蹤物體在電腦視覺上一直以來都是一個很重要的題目，本文裡我們提出一個改善CAMSHIFT的物件追蹤方法。我們使用2D histogram 包含Hue 和亮度資訊來描述目標的特徵，使用此方式能夠解決影像品質不好和具有achromatic像素點的問題。方法中並使用flooding演算法和貢獻度計算，使獲得的直方圖可以如實反映出目標的色彩資訊和各個色彩對前景和背景的辨識度。為了能適應的前景/背景的變化，我們提出更新ROI尺寸和目標直方圖的方法。實驗結果顯示，我們所提出的方法比現有的方法，能保有穩定且令人滿意的結果。

In this paper, we present an improved version of CAMSHIFT algorithm. We use a 2D histogram including hue and brightness to describe the color feature of the target. In this way, videos with poor quality or achromatic points can be better characterized. The Flooding process and contribution evaluation are used to obtain a precise target histogram which reflects true color information and discrimination ability. To be adaptive to the foreground/background variation, formula for updating ROI size and target histogram are proposed. The proposed method is compared with existing methods and shows steady and satisfactory results.

目錄
第一章 緒論 1
1.1 研究動機與目的	1
1.2 論文架構 4
第二章 相關研究 5
2.1 顏色特徵 5
2.2 CAMSHIFT演算法 7
2.3 Kernel-based object tracking演算法 11
第三章 研究方法 14
3.1 The Region of interest 14
3.2 色彩空間 15
3.3 目標物的長條圖建立 16
3.3.1 2D HY-Histogram 17
3.3.2 Floodfill 19
3.3.3 Contribution Histogram 21
3.4 Object Tracking 23
3.4.1 更新目標物的機率特徵 27
3.4.2 更新視窗大小	27
第四章 實驗結果與分析 29
4.1 與其他方法比較 30
4.2 Floodfill的影響 34
4.3 初始圈選ROI的強健性 35
4.4 追蹤失敗的例子 37
第五章 結論與未來研究方向 38
參考文獻 39
附錄：英文論文 41

圖目錄
圖1. 平均位移向量	2
圖2. HSV模型 6
圖3. 特徵空間和背投影 8
圖4. 背投影流程圖 8
圖5. 設定初始搜尋區域示意圖	8
圖6. 更新質心流程	9
圖7. 建立初始畫格的目標物顏色特徵流程圖 14
圖8. 目標和背景示意圖 15
圖9. 亮度比較圖 16
圖10. 2D直方圖 18
圖11. 水平和垂直平滑化 19
圖12. Floodfill結果比較 21
圖13. Back projection with and without contribution histogram 23
圖14. ROI and small_ROI 25
圖15. 完整流程 26
圖16. 水平垂直累積直方示意圖 28
圖17. Video_1實驗結果，其中(a)為共同的初始ROI，(1) Ours (2) CAMSHIFT (3) Kernel-based object tracking 31
圖18. Video_2實驗結果，其中(a)為共同的初始ROI，而 (1) Ours (2) CAMSHIFT(3) Kernel-based object tracking 32
圖19. Video_3實驗結果，其中(a)為共同的初始ROI，而 (1) Ours (2) CAMSHIFT (3) Kernel-based object tracking 33
圖20. Floodfill測試 (a)為初始的ROI以及所產生的mask；第二列為採用Floodfill 以改進目標物顏色特徵所得的實驗結果；第三列則沒有採用Floodfill，直接取(a)的ROI作為目標物顏色特徵所得的實驗結果 35
圖21. 框選大小不同視窗的實驗結果 36
圖22. 追蹤失敗 37

[1] K. Fukunaga and L.D. Hostetler, “The Estimation of The Gradient of A Density Function, with Applications in Pattern Recognition,” IEEE Transactions on Information Theory, vol. 21, no. 1, 1975, pp. 32-40.
[2] D. Comaniciu and P. Meer, “Mean Shift: A Robust Approach Towards Feature Space Analysis,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 24, no. 5, 2002, pp. 603-619. 
[3] D. Comaniciu, V. Ramesh, and P. Meer, “Real-time Tracking of Non-rigid Objects Using Mean Shift,” Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, 2000, pp. 142-149.
[4] D. Comaniciu, V. Ramesh, and P. Meer, “Kernel-Based Object Tracking,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25, no. 5, 2003, pp. 564-577.
[5] G.R. Bradski, “Computer Vision Face Tracking for Use in A Perceptual User Interface,” Intel Technology Journal, 2nd Quarter, 1998, pp. 13-27.
[6] The Open Source Computer Vision Library, 2001.
[7] T. Lindeberg, “Feature Detection with Automatic Scale Selection,” International Journal of Computer Vision, vol. 30, 1998, pp. 79-116.
[8] A. Bhattacharyya, “On a Measure of Divergence Between Two Statistical Populations Defined by Probability Distributions,” Bulletin of the Calcutta Mathematical Society, 1943, pp. 99-109.
[9] Y.- N. Dai, Human Tracking Algorithm Based on Multiple Features for Mobile Robot Following. Master thesis of Computer Science and Information Engineering department, National Dong Hwa University, 2009. (融合多特徵之人物追蹤演算法─應用於移動機器人之跟隨)，戴佑安，98年7月國立東華大學資訊工程學系碩士論文)
[10] R. Collins, Y. Liu & M. Leordeanu, “Online Selection of Discriminative Tracking features,” IEEE Journal of Pattern Analysis and Machine Intelligence, vol. 27, no. 10, 2005, pp.1631-1643.
[11] J.G. Allen, R.Y.D. Xu, and J.S. Jin, “Object Tracking Using CamShift Algorithm and Multiple Quantized Feature Spaces,” Proceedings of the Pan-Sydney area workshop on Visual information processing, Australian Computer Society, Inc., 2004, pp. 3-7.
[12] C.- W. Lin, Application of Mean Shift to Real-Time Visual Tracking for a Deformable Object.Master thesis of Mechanical and Electro-Mechanical Engineering department, National Sun Yat-Sen University, 2009. (應用平均位移法於變形物體之即時視覺追循，林家瑋，98年7月國立中山大學機械與機電工程學系碩士論文)