數位相機由於具有自動對焦、自動曝光，以及自動白平衡的功能，使其在操作上較傳統相機簡單而深受消費者的喜愛，此三種自動化處理對影像的品質有非常大的影響，本篇論文研究重點即在數位相機上的自動白平衡演算法。現有白平衡演算法經常使用SFU 提供的實驗影像來進行性能的比較，此組實驗影像因考量到控制亮度以避免影像的三色彩頻道發生飽和現象，因而造成影像偏暗，與數位相機所處理的影像不盡相同。本篇論文主要針對SFU實驗影像圖庫進行亮度調整，研究不同亮度環境下各種白平衡演算法之性能變化與優缺點，實驗結果顯示，當實驗影像亮度調高時，白平衡演算法之誤差亦同時增加，各種不同白平衡方法的增加幅度亦有差異，因此單獨使用此組實驗影像並不足以判斷白平衡演算法之優劣。同時論文中也提出一些改良的白平衡演算法，其一為加入空間平滑濾波器與排序標準差設定權重方式，改良標準差權重灰界理論演算法；另提出一以簡單方式混合二種白平衡演算法估計光源的成功例子，經過亮度調整影像測試後皆較前人的白平衡演算法誤差值低。

As digital cameras are equipped with useful functions such as auto focus, auto exposure and auto white balance, their relative ease of operation in comparison with traditional cameras have won over the hearts of countless consumers. The three functions featured on DSC (digital still camera) have tremendous impact on the image quality of captured images. The focal point of this study is centered on the auto white balance algorithms on digital cameras. Currently auto white balance algorithms usually use the test images provided by SFU for performance comparison. In this image set, brightness control prevents the three primary color channels from saturation. As a result, these test images are darker and substantially different from the images processed by DSCs. In this paper, brightness adjustments are made for images in the SFU test image database in an attempt to investigate performance variations of various auto white balance algorithms under different brightness conditions along with their pros and cons. Results showed that the performance of white balance algorithms deteriorates when the brightness of test images increases. Different white balancing methods also demonstrated different extents of deterioration. It is therefore apparent that using this set of test images alone is insufficient to determine the performance of white balance algorithms. Also, improved white balance algorithms are proposed in this dissertation. One method is to improve the standard deviation weighted gray world method by adding a spatial domain mean filter along with use a different weighting method. Another method combines the results of two white balance algorithms via a simple step to estimate the light source. Both methods produce lower errors than previous white balance algorithms when tested using SFU image sets with adjusted brightness.

中文摘要i
英文摘要iii
誌謝iv
目錄v
表目錄vii
圖目錄viii
第一章 緒論1
1.1 研究目的1
1.2 恆常性與白平衡1
1.2.1 色彩恆常性2
1.2.2 色溫2
1.2.3 von Kries模型4
1.3 白平衡演算法的影像5
1.4 白平衡演算法的誤差5
1.5 論文結構6
第二章 前人的自動白平衡演算法7
2.1 灰界理論(Gray World)7
2.2 最大RGB值演算法(MAX-RGB)9
2.3 百分之五最大RGB值演算法(MAX-RGB-5％)11
2.4 最大色彩值演算法(MAX-number)13
2.5 標準差權重灰界理論(SDWGW)15
2.6 邊緣灰界理論假說(Gray-edge)18
2.7 二階邊緣灰界理論假說(2nd order Gray-edge)20
2.8 閔可夫斯基範數灰界理論(Shade of Gray World)21
2.9 使用空間平滑濾波器之閔可夫斯基範數灰界理論
(general Gray World)23
2.10 實驗結果與比較25
第三章 提出的自動白平衡演算法26
3.1修正標準差權重灰界理論演算法(MSDWGW)26
3.2最大值灰界理論(Max-Gray World)混合二階邊緣灰界理論假說
(2nd order Gray-edge)37
3.3實驗結果與比較43
第四章 亮度變化對自動白平衡的影響44
4.1自行亮度調整SFU 提供的影像45
4.2使用自行亮度調整SFU 影像之實驗結果47
4.2.1 灰界理論(Gray World)47
4.2.2 最大RGB值演算法(MAX-RGB)51
4.2.3 百分之五最大RGB值演算法(MAX-RGB-5％)55
4.2.4 最大色彩值演算法(MAX-number)59
4.2.5 標準差權重灰界理論(SDWGW)63
4.2.6 邊緣灰界理論假說(Gray-edge)67
4.2.7 二階邊緣灰界理論假說(2nd order Gray-edge)71
4.2.8 閔可夫斯基範數灰界理論(Shade of Gray World)75
4.2.9 使用空間平滑濾波器之閔可夫斯基範數灰界理論
(general Gray World)79
4.2.10修正標準差權重灰界理論演算法(MSDWGW)83
4.2.11最大值灰界理論混合二階邊緣灰界理論假說87
4.3 實驗結果與比較91
第五章 結論與建議94
參考文獻96
附錄一 演算法運算的時間98
附錄二 演算法執行的程式99

[1]陳炳旭，「數位相機自動白平衡：結合灰界理論與色域對應之研究」，世新大學，碩士論文，民國九十二年
[2]林殷旭，「自動白平衡之軟硬體共同設計」，台北科技大學，碩士論文，民國九十五年
[3] M. D. Fairchild, (1998), Color Appearance Models, Addision-Wesley.
[4]香港太空館與香港大學物理系
[5]http://hometown.aol.com/lklstars5/9b.htm
[6]林宗彥，「改良式灰界理論自動白平衡演算法與應用」，國立彰化師範大學，碩士論文，民國九十四年
[7] G. Wakefield, Color Temperature and a Modern Meter, British Journal of Photography , Vol.135, pp.10-11, 1988.
[8]維基百科http://zh.wikipedia.org/wiki/%E8%89%B2%E6%B8%A9
[9]工研院http://www.cpmah.org.tw/2002/home/c91c036/www/science.htm
[10]Kobus Barnard, Lindsay Martin, Brian Funt, and Adam Coath, "A Data Set for Color Research," Color: Research and Application, Vol. 27, No. 3, pp. 148-152, 2002.
[11] Kobus Barnard, Vlad Cardei, and Brian Funt, “A Comparison of Computational Color Constancy Algorithms; Part Two: Experiments with Image Data”, IEEE Transactions in Image Processing , Vol. 11, No. 9, pp 985-996, Sept 2002.
[12]Tzan-Sheng Chiou, "Automatic White Balance for Digital Still Camera ", MSc thesis, National Taiwan University, Department of Computer Science and Information Engineering (2001)
[13]沈俊宏，「數位相機上的自動對焦與自動白平衡演算法」，國立臺灣大學，碩士論文，民國九十五年
[14]Kobus Barnard, Vlad Cardei, and Brian Funt, “A Comparison of Computational Color Constancy Algorithms; Part One: Methodology and Experiments with Synthesized Data”, IEEE Transactions in Image Processing, Vol. 11, No. 9, pp. 972-984, Sept. 2002
[15]E. H. Land and J. J. McCann, "Lightness and retinex theory," J Opt Soc Am, Vol. 61, pp. 1-11, 1971
[16]林宗漢，「使用類神經網路偵測影像色偏及模糊系統修正影像側偏之研究」，國立交通大學，碩士論文，民國九十四。
[17]H. K. Lam, O. C. Au, and C. W. Wong, "Automatic white balancing using standard deviation of RGB components,"in proc. IEEE Conf：Circuits and Systems, 2004, Vol. 3, pp. 921-924.
[18]J. van de Weijer and T. Gevers, "Color Constancy based on the Gray-Edge Hypothesis," IEEE International Conference on Image Processing, Vol. 2, 2005.
[19]G. D. Finlayson and E. Trezzi, "Shades of gray and colour constancy," IS&T/SID Twelfth Color Imageing Conference, pp. 37–41, 2004.
[20]J. van de Weijer, Th. Gevers, A. Gijsenij, "Edge-Based Color Constancy," IEEE Transactions on Image Processing, Vol. 16, pp.2207~2214, September 2007
[21]Y. C. Liu, W. H. Chan, and Y. Q. Chen, ”Automatic white balance for digital still camera ,” IEEE Trans. Consumer Electronics,vol. 41, no. 3, pp. 460-466, Aug. 1995.
[22]鍾國亮，影像處裡與電腦視覺，三版，台北，台灣東華，民國九十五年。