惡劣的天氣下光學影像極易受到不穩定的氣候所影響，例如下雨、下雪、起霧或是霾害，該種種不佳的氣候會使得原輸入影像品質因此降低。而本篇論文提出了一種新方法進行影像除霧，並且使用單影像實作在多變的環境中，不再受限於傳統的雙影像系統或是硬體儀器的輔助。該系統不只能自動調整參數，使得輸出影像的飽和度達到更好的效果外，更修改了原Retinex演算法架構，採用了影像積分的方法，提升了總體運算的速度。
光源的傳遞若被霧氣與塵霾粒子色散，而之後所接收到的影像，勢必產生模糊不清的現象，而該影像很難被拿來擷取後進行相關的電腦視覺應用，如移動偵測、追蹤與辨識等。故本篇論文提出了一種新的技術進行影像除霧。據研究，目前所知的除霧技術，主要分類成兩種架構，第一種架構使用影像回復的觀念，另一種架構則利用影像增強的原理進行操作。影像回復的架構是參考光源衰變方程式，並且試圖設法的求出該影像的深度與大氣光參數，利用兩參數去回復沒有霧氣的影像。而另一種架構則是利用影像增強的方法，去增強影像的飽和度、對比度、彩度和邊緣銳化程度。
本論文是採用了影像增強方法進行除霧，而參考的觀念是由色彩恆常性中衍生而出的Retinex 演算法，我們使用了當中的MSR演算法進行影像除霧，並修改了演算的運算架構，使用影像積分的方式達到高速運算的效果，除此之外論文使用模糊理論去自動分配參數的加權比例，進而提升飽和度的顯示。
在實驗測試中，由實驗結果我們歸類成三類，分為效果佳、效果不明顯及效果不佳。在效果佳的測試組中，又細分成影像景深的遠、中、近(室內、室外) 等測試，並且在最後會列出各演算法的比較，以此證明本論文的除霧系統在速度與最佳化參數都有不錯的表現。

Optical image is easily affected by poor weather as rain, snow, fog and haze, and this kind of poor weather causes input image low quality. Therefore, we propose a new approach for defog image, that is, it can work in a single image, instead of multi-image or hardware-assisted. Besides, the proposed method not only automatically assigns the parameters to get higher saturation value but also modify the Retinex algorithm by image integral to improve the system speed.
When receiving the object light that is scattered by droplets or haze, it is definite that received image will appear blurry. The scattered light is difficult to use in computer vision such as motion detection, tracking and recognition…etc. Therefore, we propose a new approach with defog image. According to the research, currently known defog technologies mainly classified into two types, the first type uses the concept of image restoration, and the other is the use of image enhancement. The image restoration refers to a physical model, and mainly used to figure out the depth and airlight parameters of the physical model; the other image enhancement mainly improve the saturation, contrast, color and sharpness of the edges.
This thesis is using image enhancement to defog image, and refer to Retinex algorithm that is constructed by color constancy concept. We use the MSR(Multiple-Scale Retinex) to implement defog effect, and to speed up the compute time, we use integral image to achieve. Moreover, we use Fuzzy theory to automatically assign weighted parameters, and it helps to enhance the display of saturation value.
In experimental result, we classified outcomes into three types, good, normal and bad. In the good type, it is subdivided into far, medium and near (indoor and outdoor) etc... Finally, we list the comparison of defog algorithms to prove this thesis has good performance in both calculation time and optimal parameters

[1] F. Cozman and E. Krotkov, “Depth from scattering,” in Proceedings of 1997 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Juan, Washington, pp. 801-806, 1997.
[2] E. H. Land and J. McCann, “Lightness and retinex theory,” Journal of the Optical Society of America, vol. 61, no. 1, pp.1-11, 1971.
[3] S. G. Narasimhan and S. K. Nayar, “Interactive (de) weathering of an image using physical models,” in Proceedings of 2003 IEEE International Conference on Computer Vision Workshop on Color and Photometric Methods in Computer Vision, New York, USA, pp.1-8, 2003.
[4] R. T. Tan, “Visibility in bad weather from a single image,” in Proceedings of 2008 in Computer Vision and Pattern Recognition, Anchorage, AK, pp. 1-8, 2008.
[5] R. Fattal, “Single image dehazing,” in ACM Transactions on Graphics (TOG), vol. 27, no. 3, pp.72, 2008.
[6] K. He, J. Sun, and X. Tang, “Single image haze removal using dark channel prior,” in Proceedings of 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami, FL, pp. 1956 - 1963, 2009.
[7] A. Levin, D. Lischinski, and Y. Weiss, “A closed-form solution to natural image matting,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 30, no 2, pp. 228-242, 2008.
[8] M.-J. Seow and V. K. Asari, “Ratio rule and homomorphic filter for enhancement of digital colour image,” Neurocomputing, vol. 69, no 7, pp. 954-958, 2006.
[9] B. Eriksson, “Automatic image de-weathering using curvelet-based vanishing point detection,” http://homepages.cae.wisc.edu/~beriksso/cs766.pdf. , 2010.
[10] E. H. Land, “An alternative technique for the computation of the designator in the retinex theory of color vision,” Proceedings of the National Academy of Sciences, vol. 83, no. 10, pp. 3078-3080, 1986.
[11] Z.-u. Rahman, D. J. Jobson, and G. A. Woodell, “Retinex processing for automatic image enhancement,” Journal of Electronic Imaging, vol. 13, no. 1, pp. 100-110, 2004.
[12] Y. Takematsu, T. Nakaguchi, N. Tsumura, and Y. Miyake, “Improvement of image quality using Retinex theory based on the statistical image evaluation,” Journal of The Society of Photographic Science and Technology of Japan, vol. 67, no. 4, pp. 410-416, 2004.
[13] M. Sonka, V. Hlavac, and R. Boyle, Image processing, analysis, and machine vision, Cengage Learning, 2014.
[14] A. C. Hurlbert, “The computation of color,” The Defense Technical Information Center Document, 1989.
[15] P. Viola and M. J. Jones, “Robust real-time face detection,” International journal of computer vision, vol. 57, no 2, pp. 137-154, 2004.
[16] L. A. Zadeh, “Fuzzy sets,” Information and control, vol. 8, no. 3, pp. 338-353, 1965.
[17] S. K. Nayar and S. G. Narasimhan, "Vision in bad weather," in Proceedings of 1999 the Seventh IEEE International Conference on Computer Vision, Kerkyra, pp. 820-827, 1999.
[18] S. G. Narasimhan and S. K. Nayar, “Contrast restoration of weather degraded images,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25, no. 6, pp. 713-724, 2003.