臺灣博碩士論文加值系統

現在的影像，可能因拍攝過程中各項環境因素造成視覺上的影響，其中下雨所帶來影響是本論文處理的主要問題。本論文旨在去除雨紋對於視覺上的影響，所以我們的第一個問題是在影像中找到雨滴，第二個問題則是去除雨紋。雨滴是由上往下掉落形成了雨紋，因此利用梯度偵測其垂直紋理就能找到可能的目標。接著我們將原始畫面通過低通濾波，得到低頻影像，並用原始影像減去低頻影像得到差異影像。將差異影像利用DCT（Discrete Cosine Transform）從空間域轉換至頻域，再於頻域中抑制相關的垂直紋理，藉此去除雨紋。
另一方面，因為去除雨滴在影像中帶來的影響需要一定程度的運算能力，所以目前去雨的研究當中，同時符合動態場景以及快速運算的去雨研究並不多。本論文利用CUDA（Compute Unified Device Architecture）加速運算，不使用前後畫面的資訊，達到單張影像單次處理的結果，期待能加強影像處理的速度，使本研究滿足動態畫面以及應用的需求。

The video image which we filmed now may affect by lots of environment factors. In this thesis, we focus on the effect of rain-drops in an image. Remove the visual influence caused by rain-drop is the main part in the thesis. The rain-drops in the image are found out at first, and then the rain-drops are removed by our algorithm. Water drops from top to bottom made rain-drops, so the gradient is used to detect vertical texture in the image and find out the possible targets. Process the source video with the low-pass filter. Subtract processed image from the source image and make the difference of two images as a discrepant image. Turn discrepant image from spatial domain to frequency domain by DCT (Discrete Cosine Transform). Then we reduce the vertical texture factors in the frequency domain, to decrease the effect of rain-drops.
In the other hand, to remove the visual influence caused by rain-drop needs well calculating machine’s performance, only a few researches mention that. In this thesis, we use CUDA (Compute Unified Device Architecture) to speed up the algorithm processing. One image is processed in one time, there is no sequence information of image is used. We except that approach will speed up the algorithm processing and make remove the visual influence caused by rain-drop in dynamic scene and processing can be implemented.

摘 要 i
ABSTRACT ii
誌 謝 iv
目 錄 v
表目錄 vii
圖目錄 viii
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機 2
1.3 文獻回顧 3
1.4 論文架構 5
第二章 相關知識介紹 6
2.1 色彩空間 6
2.2 雨紋特性分析 8
2.3 Sobel梯度偵側演算法 9
2.4 OpenCV介紹 11
2.5 GPU簡介 11
2.5.1 CUDA 軟體架構說明 12
2.5.2 CUDA 硬體架構說明 14
第三章 研究方法 18
3.1 系統架構 18
3.2 輸入影像與色彩空間轉換 22
3.3 雨紋特性分析 22
3.4 垂直紋理偵側 23
3.5 雜點去除與膨脹、侵蝕 24
3.6 影像低通濾波 28
3.7 差異影像與高頻成分 30
3.8 DCT與IDCT 31
3.9 適應性邊緣銳化 37
3.10 HSV to RGB 39
第四章 實驗結果與分析 40
4.1 雨紋測試影片簡介 40
4.2 靜態去雨分析 40
4.3 動態去雨分析 42
4.4 人造雨去雨分析 44
4.5 CUDA加速 46
4.5.1六次低通濾波加速 47
4.5.2 Sobel梯度偵測加速 48
4.5.3 DCT&IDCT加速 48
第五章 結論與未來工作 51
5.1 結論 51
5.2 未來工作 51
5.2.1 CUDA加速（快速DCT&IDCT） 51
5.2.2 8×8 DCT轉換物理特性 52
5.2.3 以Motion Vector為基礎之去雨演算法 52
參考文獻 53

[1]X. Zhang, H. Li, Y. Qi, W. K. Leow, and T. K. Ng, “Rain removal in video by combining temporal and chromatic properties,” in Proc. IEEE International Conference Multimedia Expo, Toronto, Ont. Canada, Jul. 2006, pp. 461–464.
[2]K. Garg and S. K. Nayar, “Detection and removal of rain from videos,” in Proc. IEEE Conference Computer Vision and Pattern Recognition, vol. 1, Jun. 2004, pp. 528–535.
[3]K. Garg and S. K. Nayar, “When does a camera see rain?,” in Proc. IEEE International. Conference Computer Vision, vol. 2, Oct. 2005, pp. 1067-1074
[4]M. Zhou, Z. Zhu, R. Deng, and S. Fang, “Rain detection and removal of sequential images,” in Proc. Chinese Control and Decision Conference (CCDC), Mianyang, May 2011, pp. 615-618.
[5]J. Xu, W. Zhao, P. Liu and X. Tang, “Removing rain and snow in a single image using guided filter,” in Proc. Computer Science and Automation Engineering (CSAE), May 2012, pp.304-307
[6]S.-C. Pei, Y.-T. Tsai, and C.-Y. Lee,“Removing rain and snow in a single image using saturation and visibility features,” in Proc. IEEE International Conference On Multimedia and Expo Workshops (ICMEW), Chengdu, Jul. 2014, pp.1-6.
[7]C.-H. Tan, J. Chen, and L.-P. Chau, “Dynamic scene rain removal for moving cameras,” in Proc. International Conference Digital Signal Processing, Hong Kong, Aug. 2014, pp. 372-376.
[8]C.-C. Chen, L.-W. Kang, “Visual Depth Guided Color Image Rain Streaks Removal Using Sparse Coding,” IEEE Transactions Circuits and Systems for Video Technology , Aug. 2014, pp.1430-1455
[9]L.-W. Kang, C.-W. Lin, and Y.-H. Fu, “Automatic single-image-based rain streaks removal via image decomposition,” IEEE Transactions Image Process., vol. 21, no. 4, Apr. 2012, pp. 1742–1755.
[10]M. Shen and P. Xue, “A fast algorithm for rain detection and removal from videos,” in Proc. IEEE International Conference Multimedia Expo, Jul. 2011, pp. 1–6.
[11]J. Chen and L.-P. Chau, “A Rain Pixel Recovery Algorithm for Videos With Highly Dynamic Scenes,” IEEE Transactions on Image Processing, Mar. 2014, pp. 1097-1104.
[12]J. Bossu, N. Hautière, and J. P. Tarel, “Rain or snow detection in image sequences through use of a histogram of orientation of streaks,” International Journal of Computer Vision, vol. 93, no. 3, Jul. 2011, pp. 348–367.
[13]M. R. Kanthan, S. N. Sujatha, “Rain drop detection and removal using K-Means clustering,” in Proc. IEEE International Conference Computational Intelligence and Computing Research (ICCIC), Dec. 2015, pp.1-5
[14]A. K. Tripathi and S. Mukhopadhyay, “A probabilistic approach for detection and removal of rain from videos,” IETE Journal of Research, vol. 57, no. 1, Mar. 2011, pp. 82–91.
[15]L.-J. Kau , T.-L. Lee,“An Efficient and Self-Adapted Approach to the Sharpening of Color Images,” The Scientific World Journal, Sept. 2013, pp. 1-18
[16]陳松鴻，2013，多干擾環境下之自適應性視訊監控系統。碩士論文，國立國立臺北科技大學電機工程研究所，臺北。
[17]維基百科，HSL和HSV色彩空間。[online]. Available: https://zh.wikipedia.org/wiki/
HSL%E5%92%8CHSV%E8%89%B2%E5%BD%A9%E7%A9%BA%E9%97%B4
[18]滄者極限，GTX970。[online]. Available: http://www.coolaler.com/showthread.php/32
3681
[19]OpenCV API Reference, Miscellaneous Image Transformations,[online]Available: http://
docs.opencv.org/2.4/modules/imgproc/doc/miscellaneous_transformations.html
[20]CUDA Toolkit Documentation,Introduction, [online] Available : http://docs.nvidia.com/c
uda/cuda-c-programming-guide/index.html#thread-hierarchy__grid-of-thread-blocks
[21]CUDA Toolkit Documentation,Compute Capability Class,[online] Available : http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#compute-capability-2-x
[22]Nvidia Geforce , GTX 970 specifications, [online] Available :http://www.geforce.com.tw
/hardware/desktop-gpus/geforce-gtx-670/specifications
[23]高立人，資料壓縮, Lecture11: The DCT-based Coding Standard, JPEG, [online] Available: http://www.cc.ntut.edu.tw/~ljkau/Frame.htm