基於色彩空間的影像偵測與強化技術

Color Space-based Digital Image Detection and Enhancement

李丞祐(Chen-Yu Lee)

指導教授 ： 貝蘇章

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2013年)

https://doi.org/10.6342/NTU.2013.02092

摘要

在數位影像拍攝及處理的過程中，總是存在著許多問題。如拍攝人像時，需抓取人臉的位置以此提升鏡頭在對焦上的準確度。又如在水下攝影時，光線經過水體的衰減後，環境光的白光中，除了藍光外都有大幅度的衰減，進而使的影像失去其在白熾光下拍攝所應有的顏色。而在雨中與雪中攝影時，則會遇到雨水及雪在快門曝光的時間內，快速通過拍攝範圍內，進而造成影像中有細條狀雜訊的問題。這些問題解決時方法的好壞，會嚴重影響到運算的成本與時間，因此，個個問題都需要一套快速而有效的演算法來解決。
本篇論文中，分別探討膚色偵測、水下影像增強以及雨與雪景的移除。在膚色偵測的部分，將分別介紹在不同的色彩空間中的偵測方式[1]-[6]，並提出一套基於影像正規劃所設計的膚色偵測方式。而在水下影像增強方面則是介紹基於色彩模型[7]與波長補償模型[8]建立的水下影像處理方式，並提出了一套基於色彩轉換[17]與dark channel prior-based [10], [11]的新的處理方式。最後則是單張影像去雨與除雪，其中介紹了guided filter [19], [21], [29]的去雨及除雪的方式，並提出了一種基於頻譜分析[20]與物件辨識與移除的影像處理方式。

關鍵字

膚色偵測 ； 水下影像增強 ； 除雨 ； 除雪

並列摘要

In the procedure of capturing and processing a photo, there are always a lot of problems in the course. For example, when we want to take a picture for someone, we usually want to focus on the people. Finding the position of the faces in the photo by features such as skin color will play an important role in the procedure. At the time taking a photo in the water, we will find that the ambient light becomes blue instead of white. The intensity of the photos taken in the water will have the highest average value in the blue plate. When we take a picture in the raining or snowing day, the photographs are disturbed by the rain or snow streaks which appear in the range of the scene. It will let the photographs having a lot of steak-shape noise. The methods to solve these kinds of problems will severely affect the time consumption.
We will discuss the methods of skin-color detection, underwater image enhancement and rain/snow removal in single image in this thesis. In the part of skin color detection, we will introduce the methods of detecting the skin color in different color spaces [1]-[6]. Besides, we propose a new method for skin color detection which is based on image normalization. In the part of underwater image enhancement, we introduce two methods to fix the photograph. One of the method is based on the colour model [7] and the other is based on wavelength compensation and dehazing [8]. Besides, we also propose a method based on color transfer [17] and dark channel prior-based [10], [11]. In the part of rain/snow removal in single image, we will introduce the methods based on guided filter [19], [21], [29]. We propose a new method based on frequency analysis [20] and object recognition to solve the problem of rain and snow in the photograph.

並列關鍵字

Skin color detection ； Underwater image enhancement ； Rain/snow removal

參考文獻

[2] Phil Chen and Dr. Christos Grecos, “A fast skin region detector ,” ESC Division Research, 2005.

[3] Tarek M. Mahmoud, “A new fast skin color detection technique,” in World Academy of Science, Engineering and Technology, pp. 498-502, 2008.

[4] P. Kakumanu, S. Makrogiannis, and N. Bourbakis, “A survey of skin-color modeling and detection methods,” Pattern Recognition, vol. 40, pp. 1106-1112, 2007.

[6] Ming-Ji Zhang , Wei-Qiang Wang , Qing-Fang Zheng , Wen Gao, “Skin-color detection based on adaptive thresholds,” Proceedings of the Third International Conference on Image and Graphics, pp.250-253, December 18 - 20, 2004.

[8] Kashif Iqbal, Rosalina Abdul Salam, Azam Osman, and Abdullah Zawawi Talib, “Underwater image enhancement using an integrated colour model,” IAENG International Journal of Computer Science, 34:2, 2007.