陰影與物體分離已成為再電腦視覺應用中一個重要的課題。然而，此問題並非如此顯而易解的。一張輸入的影像可以等於物體本身影像以及陰影的相乘積。本篇提出兩個方法來分離單張的陰影與物體。首先將輸入影像跟微分過濾器作迴旋積分，而像素的微分值隨後會分類成陰影相關或者物體相關。於是在本篇假設影像的微分值只可能由物體本身的變化或者陰影的變化其中之一所引起，不可能同時發生。我們提出用權重圖的方法來分離單張彩色圖中的陰影，至於灰階影像則採另一種Adaboost方法為核心來分離影像。最後，我們可用分類結果的微分影像還原回復成我們想要的物體本身以及陰影分離圖。我們用合成影像來解說闡述方法。而在實驗結果中，也證明了我們方法亦能有效的分離現實生活中影像的陰影。

Intrinsic image extraction has long been an important problem for computer vision applications. However, this problem is nontrivial at all because it is an ill-posed problem. An input image is a product of its shading image and reflectance image. Two methods for extracting intrinsic images form a sing image is presented. The method first convolves the input image with derivative filters. The pixels of filtered images are then classified shading-related or reflectance-related. This paper assumed that the derivative is either caused by shading or reflectance, but not both. We use weighted-map method to separate one single color image and use adaboost-based method to separate a gray-scale image. Finally, the intrinsic images of input image can be reintegrated from the classification results of the filtered images. We use a synthetic image to demonstrate result. Besides, our approach also works on real image data in experimental results.

[1] H.G. Barrow and J.M. Tenebaum, “Recovering Intrinsic Scene Characteristics form Images,” Computer Vision Systems, A, Hanson and E. Riseman, eds., pp.3-26.Academic Press, 1978.
[2] S.L. Chang, L.S. Chen, Y.C. Chung, and S.W. Chen, “Automatic license plate recognition,” IEEE Transaction on Intelligent Transportation Systems, 5(1), pp. 42-54, 2004.
[3] B.V. Funt, m. S. Drew, and M. Brokington, “Recovering Shading from Color Images, ”The 2nd European Conference on Computer Vision, pp. 124-132, 1992.
[4] C.. Fang, S. W. Chen, and C.S. Fuh, “Automatic Change Detection of Driving Environments in a Vision-Based Driver Assistance System,” IEEE Transaction on Neural Networks, 14(3), pp.646-657, 2003.
[5] G.D. Finlayson and S.D. Hordley, “Color Constancy at a Pixel,” J. of the Optical Society of America, 18(2), pp. 253-264, 2001.
[6] G.D. Finlayson, S.D. Hordley, L.Cheng and M. S. Drew, “On the Removal of Shadows form Images,” IEEE Trans. on PAMI, 28(1), pp. 59-68, 2006.
[7] Y. Freund and R.E. Schapire, “A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting,” J. Computer and System Sciences, Vol. 55, No.1, pp. 119-139, 1997.
[8] D. Heeger and J. Bergen, “Pyramid-Based Texture Analysis/Synthesis,” Proc. Special Interest Group on GRAPHics and Interactive Techniques, Computer graphics, pp. 229-238, 1995.
[9] Kingdom, F.A.A., Rangwala, S. & Hammmamji, “Chromatic Properties of the Color Shading Effect,” Vision Research, 45, pp. 1425-1437, 2005.
[10] T. Leung and J. Malik, “Recognizing Surfaces Using Three-Dimensional Taxons,” IEEE International Conference on Computer Vision, pp. 1010-1017, 1999.
[11] Y. Matsushita, K. Nishino, K. Ikeuchi, and M. Sakauchi, ”Illumination Normalization with Time-Dependent Intrinsic Images for Video Surveillance,” Proc. 2003 Conference Computer Vision and Pattern Recognition, Vol. 1, pp. 3-10, 2003.
[12] B.A. Olshausen and D.J. Field, ”Emergence of Simple Cell Receptive Field Properties by Learning a Sparse Code for Natural Images,” Nature, 381: pp. 607-608, 1996.
[13] E.P. Simoncelli, “Statistical models for images: compression, restoration and synthesis,” Asilomar Conference on Signals, Systems and Computers, 1, pp. 673-678, 1997.
[14] K. Tieu and P. Viola, “Boosting Image Retrieval,” Proc IEEE Computer Vison and Pattern Recognition, Vol. 1, pp. 228-235, 2000.
[15] M.F. Tappen, W.T. Freeman, and E.H. Adelson, “Recovering Intrinsic Images from a Single Image,” Transaction on Pattern Analysis and Machine Intelligence, 27(9), pp.1459-1472, 2005.
[16] Y.T. Tsai and J.J. Lien, “Efficient Object Segmentation Using Digital Matting for MPEG Video Sequences,” Asian Conference on Computer Vision, (2), pp.591-601, 2006.
[17] Y. Weiss, “Deriving Intrinsic Images from Image Sequences,” IEEE International Conference on Computer Vision, 2, pp. 68-75, 2001.