一般數位相機所拍攝出的靜態影像，因受限於類比數位轉換器之解析度，所以單一曝光值影像所能表達的動態範圍(Dynamic Range)常遠低於人眼當時所看到的真實景象。為了解決這個問題，利用多張不同曝光值之影像來合成一張高動態範圍影像(High Dynamic Range Image ,HDRI)之演算法應運而生，此類之合成可解決因單一曝光值影像動態範圍過小而產生色彩漸層遠低於人眼所見之問題。
但在利用多張不同曝光值影像做融合之演算法中，由於取得不同曝光度數位影像時會有時間上的差異，因此多張景像內之物體會有因移動而產生殘影(Ghost)的現象，所以如何將移動物件校正及將殘影移除以得到清晰之高動態範圍影像是一個值得研究的課題。
本論文在以梯度為基礎之多曝光值高動態範圍影像合成法上，提出了一個用於多曝光值高動態範圍影像合成法中具移動物件校正與殘影移除之演算法，藉由直方圖等化處理來偵測會發生殘影的移動物件，並利用邊緣偵測來強化移動物件之輪廓，以便將移動物件校正及將殘影移除。然後再配合以梯度為基礎之多曝光值高動態範圍影像合成演算法，來產生一張高動態範圍之影像。此方法由於事先將移動物件偵測出來加以校正再作殘影之移除，因此此一合併多曝光值影像所得到之高動態範圍影像，將會有較佳之效果。
經實驗驗證，本論文所提出的演算法所合成的高動態範圍影像，由於事先將移動物件加以偵測並校正，因此在殘影移除之效果上有大幅度的提升。較之於其他殘影消除之演算法，殘影移除的部份更加清晰，而所生成之高動態範圍影像在整體上會有更佳的視覺表現。

關鍵詞：高動態範圍影像，直方圖等化，邊緣偵測，殘影，高斯混合函數。

The image captured by the common digital camera, because of the limited of single exposure image, the dynamic range is far lower than the image seen by the human eye. To resolve this issue, the algorithms synthesize images with different exposure to construct a High Dynamic Range Image, HDRI have been invented. These kind of algorithms resolve the issue of the single exposure image with insufficient dynamic range-result in the color far lower than the human eye.
However in the algorithms combining images with different exposure, it requires images with different exposure time, and because of the small time difference when capturing the images, the movements of the objects in the picture would result in the Ghost effect. Therefore the method to adjust the moving objects and to eliminate the ghost effect in order to obtain a high resolution and high dynamic range image is a interesting issue. This article uses gradient based synthesized multiple exposure time HDR image as foundation and propose different algorithms to adjust the moving objects and to eliminate the ghost effect. Using histogram maps process to sensor the moving objects, then edge detect to intensify the features of the moving objects to adjust and eliminate the ghost effect. The method would then collaborate with the algorithm of using gradient as foundation to synthesize different exposure images to produce a high dynamic range image. Because the method sensors then adjust the moving object to eliminate the ghost effect, the result of combining different images of high dynamic range image would have a better quality.
The evidence suggests that because the adjustment of the moving objects, the ghost effect have been significantly eliminated, thus the algorithm used to produce the high dynamic range image have remarkably improved the quality of the image. Comparing to other algorithms, previous parts of image suffering from the ghost effect are more clear and distinct, and also have a better visual performance in the produced high dynamic range image.

Keyword：High dynamic range image(HDRI), histogram,
edge detect,ghost,Gassian Blending Function.

[1]Tae-Hong Min*, Rae-Hong Park*, and Son-Keun Chang,
”Histogram based ghost removal in high dynamic range
images”,Proceedings of the 2009 IEEE international conference on Multimedia.

[2] W. C. Kao, C. C. Hsu, L. Y. Chen, C. C. Kao and S. H. Chen,
“Integrating image fusion and motion stabilization for capturing still images in high dynamo Electronics, vol. 52, no. ic range scenes,” IEEE Trans. Consumer3, pp. 735-741, Aug. 2006.

[3] W. C. Kao, “High Dynamic Range Imaging by Fusion Multiple
Raw Images and Tone Reproduction,” IEEE Trans. Consumer
Electronics,vol. 54, no. 1, pp. 10-15, Feb. 2008.

[4] A. Rövid, A. R. Várkonyi-Kóczy, T. Hashimoto, S. Balogh and Y.
Shimodaira, “Gradient Based Synthesized Multiple Exposure Time HDR Image,” Instrumentation and Measurement Technology Conference - IMTC 2007 Warsaw, Poland, May 1-3, 2007.

[5] A. R. Várkonyi-Kóczy, A. Rövid and T. Hashimoto,
“Gradient-Based Synthesized Multiple Exposure Time Color HDR Image,” IEEETRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 8, AUGUST 2008.

[6] H. B. Bidasaria, “Defining and rendering of textured objects
through the use of exponential functions,” Graph. Models Image Process vol.54,no. 2, pp. 97–102, 1992.

[7] F. Drago, K. Myszkowski, T. Annen and N. Chiba. “Adaptive
Logarithmic Mapping For Displaying High Contrast Scenes,”
EUROGRAPHICS . Volume 22(2003), Number 3. 2003.

[8] G. Qiu and J. Duan, “An Optimal Tone Reproduction Curve
Operator for the Display of High Dynamic Range Image,” 2005.
ISCAS 2005. IEEE International Symposium on Circuits and
Systems, 23-26 May 2005.

[9] J. Duan and G. Qiu, “Fast Tone Mapping for High Dynamic
Range Images”, ICPR 2004.Proceedings of the 17th
International Conference on Pattern Recongnition , 23-26 Aug.2004.

[10] J. Guan, G. Qiu, “Display HDR Image using a Gain Map”2007.
ICIP2007. IEEE International Conference on Image
Processing,Sept. 16 2007-Oct. 19 2007.

[11]R.A.Hummel,”Histogram modification techniques,”Compute
Graphics Images Process,4,209-224 1975.

[12]Gonzalez, R.C.and R.E Woods(2002)Digital Image
Processing,2nd Ed.,Prentice Hall,New Jersey,NJ.pp 222-224.

[13]M.Sonka,V.Hilavac,R.Boyle,Image Processing,Analysis,and
Machine Vision,PWS Publishing,1998.

[14]Georges Matheron,”Randon Sets and Integral Geome-
try”,Wiley,New York,1975.

[15]N.Ramesh,J.H.Yoo,I.K.Sethi,”Vision Image and Signal
Processing”,IEEE Proceedings-Volume:142 5,1995,pp
271-279.

[16] G. Ward, “High dynamic range imaging,” In Proceedings of the
Ninth Color Imaging Conference, November 2001.

[17] G. Ward, 1991. Real Pixels. In James Arvo Ed., Graphic
Gems II,Chapter 11.5, pp.80-83. Cambridge MA: AP
Professional.

[18] G. Ward and Anyhere Software, “High Dynamic Range Image
Encodings.”

[19] P. E. Debevec and J. Malik, “Recovering High Dynamic
Range Radiance Maps from Photographs,” SIGGRAPH ’97
Conf. Proc.pp.369-378, Aug. 1997.