文件列表:

fast_deblur (0, 2010-04-08)
fast_deblur\bin (0, 2010-04-08)
fast_deblur\bin\cv110.dll (876544, 2008-10-17)
fast_deblur\bin\cxcore110.dll (958464, 2008-10-17)
fast_deblur\bin\fast_deblur.ex_ (192512, 2010-01-24)
fast_deblur\bin\highgui110.dll (630784, 2008-10-17)
fast_deblur\bin\libfftw3f-3.dll (1382280, 2006-07-04)
fast_deblur\bin\libiomp5md.dll (530616, 2009-06-25)
fast_deblur\bin\libmmd.dll (2549944, 2009-06-25)
fast_deblur\bin\Microsoft.VC80.CRT.manifest (1870, 2009-07-11)
fast_deblur\bin\msvcm80.dll (479232, 2009-07-12)
fast_deblur\bin\msvcp80.dll (554832, 2009-07-12)
fast_deblur\bin\msvcr80.dll (632656, 2009-07-12)
fast_deblur\bin\vcomp.dll (65536, 2009-07-11)
fast_deblur\ChangeLog.txt (408, 2010-04-08)
fast_deblur\examples (0, 2010-04-08)
fast_deblur\examples\boy_statue.jpg (339781, 2009-01-14)
fast_deblur\examples\fishes.jpg (59141, 2009-09-11)
fast_deblur\examples\hanzi.jpg (537411, 2009-12-29)
fast_deblur\examples\harubang.jpg (69339, 2009-09-11)
fast_deblur\examples\hollywood.jpg (647770, 2009-12-29)
fast_deblur\examples\run_deblur.ba_ (436, 2010-01-24)
fast_deblur\examples\summerhouse.jpg (551197, 2009-12-29)
fast_deblur\examples\Thumbs.db (59904, 2009-12-30)

Fast Motion Deblurring v0.1 – README http://cg.postech.ac.kr/research/fast_motion_deblurring Sunghyun Cho http://home.postech.ac.kr/~sodomau sodomau@postech.ac.kr Dec. 29, 2009 ________________ IMPORTANT NOTE ________________ Usage of the executable file is permitted only for academic purpose. Please do not redistribute the executable. If you generate data (e.g., images, tables of processing times, etc.) using the executable for an academic publication, please include the following citation in your paper. @article{SCho_deblur_2009, title={Fast Motion Deblurring}, author={Sunghyun Cho and Seuongyong Lee}, journal={ACM Transactions on Graphics (SIGGRAPH ASIA 2009)}, year={2009}, volume={28}, number={5}, pages={article no. 145} } If you want to use the executable for other purposes, please contact us. Note that the executable provided here is a CPU version not a GPU-accelerated version, although we also implemented a GPU-accelerated deblurring program and used it in our paper. ________________ HOW TO INSTALL ________________ The executable was compiled using Microsoft Visual C++ 2005 SP1 and Microsoft Windows XP 32 bit version. We did not test any other environments than Windows XP 32bit, and we do not guarantee that our executable will work on other environments. Since some e-mail service providers (e.g., gmail) do not allow sending executable files via e-mail, we changed the extensions of some files in the zip file. So, please change the extensions of the following files: * bin\fast_deblur.ex_ --> bin\fast_deblur.exe * examples\run_deblur.ba_ --> examples\run_deblur.bat Then, you can run the exe file or the bat file without any additional setup process. ________________ HOW TO USE ________________ You may run the executable on a Windows terminal program "cmd.exe" as follows: C:\FastDeblur> fastdeblur.exe input.png output.png 35 35 0.5 0.001 Y * input.png: The first parameter is a path of an input image file. Supported file formats: jpeg, bmp, png, etc. * output.png: The second parameter is a path of the output image file. The output blur kernel is saved as {output_file_path}.psf.png * 35 35: The third and fourth parameters are the width and height of a blur kernel. Set the width and height larger than the exact size of a blur kernel. For example, if the exact size of a blur kernel is 15x15, then you'd better set the size parameters 21x21, 25x25, or something like these. By doing this, translation of an estimated blur kernel can be handled safely. * 0.5: The fifth parameter is a range sigma for bilateral filtering in the prediction step. 0.5 works for most cases. If an input image has many saturated pixels, larger values may help. * 0.001: The sixth parameter is regularization strength for the final deconvolution. A larger value produces an image with less noise. * Y: The seventh parameter is about whether you use gamma correction or not. 'Y' means that you use gamma correction, and 'N' means that you do not use it. Most images are gamma encoded, so 'Y' should be used to linearize pixel intensities during the deblurring process. If an input image is a synthetically blurred one, then 'N' should be used. Because we have improved and modified many parts of our source code after the submission of the camera-ready version of our paper, the speed is higher than the reported processing times in the paper. Also, the resulting quality may be slightly different from the examples of the paper. However, we did not change any algorithmic details. Please note that the algorithm is still not perfectly stable and sometimes may converge to an incorrect result. When you obtain such an incorrect result, please re-try to deblur with a slightly changed parameters (e.g., changing blur kernel sizes by 1 pixel). For some results in our paper, we used a complex ringing suppression method of [Shan et al. 2008], which is described in the authors' document about additional programming details: http://www.cse.cuhk.edu.hk/~leojia/projects/motion_deblurring/implementation.pdf. However, the executable does not include this functionality. We also included a batch file in the directory 'examples'. The batch file runs the deblurring executable for producing deblurred results of example images. Please note that parameter values in the batch file are not carefully tuned. For algorithmic details, please refer to our paper.

近期下载者：

相关文件：

评论：[我要评论] [举报此文件]

收藏者：