5.1. Conclusion
In this thesis, we modify Fattal et al. algorithm which is originally used in the tone-reproduction for the high dynamic-range image, and if can also be applied to the shadow removal. The original method doesn’t produce the satisfying results because of the blurring affect, so we first detect the fuzzy edges by the multi-resolution approach, and then subtract the local means of the gradients within those regions. This modification reduces the blurring affect very well, but there still have many improvements for us to work out. How do we adjust the gradients within the fuzzy edge? Is it suitable to simply subtract the local-mean without the consideration of the noise? The fuzzy edge detection is also the important step which can influence the performance severely. In our knowledge of the “fuzzy edge” detection, the way is using the multi-resolution to detect the possible fuzzy edges. Besides, we use the “binary” edge detection approach to decide whether the specific position in the image is fuzzy edge or not, the approach may be extends to the “fuzzy” edge detection by applying the fuzzy set. However, we can not ensure affirmly whether the detection method can be used in the generalized case, and whether the percentage of the correct fuzzy edge detection fulfills our requirements? So, we still have more experiments to do to confirm the correctness of the detection method. The
“color shift” is another challenge in the shadow removal. It’s not enough to deal with the shadowing images by only considering the intensity information. Instead, the color information should be included in the consideration in order to correct the problem of “color shift“.
5.2. Future work
We adopt the concept of gradient domain to deal with the shadow removal problem. The concept of the attenuation in gradient domain can also be migrated to the wavelet domain because the HH, LH, HL sub-bands after the high-pass filter also represent the higher frequency of the original image, which is the same as the gradient indicates. Maybe we can apply our shadow removal method to the wavelet domain. In addition, our shadow removal method still leaks the chromatic information, which causes the problem of “color shift“. We have found that the illumination-invariant approach may be overcome this problem, however, due to the limit of time, we expect to correct the color shift in the future. If the chromatic information is included, not only fuzzy edges but the sharp edges can also be detected correctly, this enhancement will then fulfill the requirement of the shadow removal.
References
[1] Stephen E. Palmer, Vision science: photons to phenomenology, MIT Press, Cambridge, 1999.
[2] Debevec, P. E. and Malik, J., “Recovering high dynamic range radiance maps from photographs”, In Proc. ACM SIGGRAPH, pp.369–378, 1997.
[3] T. Mitsunaga and S. K. Nayar, “Radiometric Self Calibration”, Proc. of IEEE Conference on CVPR, Fort Collins, June 1999.
[4] M. D. Grossberg and S. K. Nayar, “High Dynamic Range from Multiple Images:
Which Exposures to Combine?”, In Proc. ICCV Workshop on Color and Photometric Methods in Computer Vision, Nice, France, October 2003.
[5] S. K. Nayar and T. Mitsunaga, “High Dynamic Range Imaging: Spatially Varying Pixel Exposures”, Proc. of IEEE Conference on CVPR, Hilton Head Island, South Carolina, June 2000.
[6] S. K. Nayar and V. Branzoi, “Adaptive Dynamic Range Imaging: Optical Control of Pixel Exposures Over Space and Time”, In Proc. ICCV, Nice, France, October 2003.
[7] Press, W. H., Teukolsky, S. A., Vetterling, W. T., and Flannery, B. P., Numerical Recipes in C: The Art of Scientific Computing, 2nd edition, Cambridge University Press, 1992.
[8] R. Fattal, D. Lischinski, and M.Werman, “Gradient domain high dynamic range compression”, Proc. of SIGGRAPH, pp.249–356, July 2002.
[9] Tumblin, J., and Rushmeier, H. E., “Tone reproduction for realistic images”, IEEE Computer Graphics and Applications 13, 6, pp.42–48, November 1993.
[10] M.F. Tappen, W.T. Freeman, and E.H. Adelson, “Recovering Intrinsic Images from a Single Image”, Proc. of Neural Information Processing Systems, 2002.
[11] Y. Weiss, “Deriving intrinsic images from image sequences”, In Proc. ICCV, Vol. 2, pp.68-75, 2001.
[12] J.M. Rubin and W.A. Richards, “Color vision and image intensities: When are changes material?” , Biol. Cybern., vol. 45, no. 3, pp.215–226, 1982.
[13] Perona, P., and Malik, J., “Scale-space and edge detection using anisotropic diffusion.”, IEEE Transactions on PAMI 12, 7 , pp.629–639, July 1990
[14] Ruifeng Xu, and Sumanta N. Pattanaik, “High Dynamic Range Image Display Using Level Set Framework.”, 2003
[15] Tumblin, J., and Turk, G., “LCIS: A boundary hierarchy for detail-preserving contrast reduction”, In Proc. ACM SIGGRAPH, pp.83–90, 1999.
[16] Jobson, D. J., Rahman, Z., and Woodell, G. A., “A multi-scale Retinex for bridging the gap between color images and the human observation of scenes”, IEEE
Transactions on Image Processing 6, 7, pp.965–976, July 1997
[17] Frédo Durand, and Julie Dorsey, “Fast bilateral filtering for the display of high dynamic range images”, ACM Transactions on Graphics, v.21 n.3, July 2002 [18] Choudhury P, and Tumblin J., “The trilateral filter for high contrast images and
meshes”, In Proceedings of the Eurographics Symposium on Rendering, pp.186–96, 2003
[19] Stockham, J. T. G., “Image processing in the context of a visual model.”, In Proc. of the IEEE, vol. 60, pp.828–842, 1972.
[20] Pattanaik, S. N., Ferwerda, J. A., Fairchild, M. D., and Greenberg, D. P., “A multi-scale model of adaptation and spatial vision for realistic image display.”, In Proc. ACM SIGGRAPH, pp.287–298, 1998.
[21] Dicarlo, J. M., and Wandell, B. A., “Rendering high dynamic range images”, In Proc.
of the SPIE: Image Sensors, vol. 3965, pp.392–401, 2001
[22] Masashi Baba and Naoki Asada, “Shadow removal from a real picture”, Proc. of the SIGGRAPH conference on Sketches & applications: in conjunction with the 30th annual conference on Computer graphics and interactive techniques, pp.27-31, San Diego, California, July 2003
[23] A. Koschan and M. Abidi, “Detection and Classification of Edges in Color Images”, Signal Processing Magazine, Special Issue on Color Image Processing, Vol. 22, No. 1, pp.64-73, 2005
[24] Graham D. Finlayson, Mark S. Drew, and Cheng Lu, “Intrinsic Images by Entropy Minimization”, ECCV, Vol. 3023, pp. 582-595, 2004
[25] Y. C. Chung, J. M. Wang, R. R. Bailey, S. W. Chen, S. L. Chang, and S. Cherng,
“Physics-based Extraction of Intrinsic Images from a Single Image”, 17th ICPR, Vol.
4, pp.693-696, August 2004.
[26] Rafael C. Gonzalez, and Richard E. Woods, Digital Image Processing, 2nd edition, Prentice Hall press, 2002.
[27] Ashikhmin, M., “A Tone Mapping Algorithm for High Contrast Images”, The Proc.
of 13th Eurographics Workshop on Rendering, pp.145-155, 2002.
[28] J. M. Geusebroek, R. van den Boomgaard, A. W. M. Smeulders, and H. Geerts,
“Color invariance.”, IEEE Transactions on PAMI, vol. 23, no. 12, pp.1338–1350, 2001.