CSE logo University of Washington Computer Science & Engineering
  CSE Home   About Us    Search    Contact Info 

Spatially variant non-blind deconvolution

 
[Click to download a package with the executable and examples]
[More obscure glass filtered images and kernels]
[More spatially variant blurred images and kernels]
 
There are two command-line executables in the package.

spvdeconv.exe : It is for spatially variant non-blind image deconvolution. It takes a blurred image and a spatially variant blur kernel (spvkernel). The command is:
spvdeconv.exe distort_image_name out_image_name kernel_file_name smoothness_weight

lockernel2spvkernel.exe: It is used to transform location kernels into a spvkernel

For more details about the format of a sparse kernel file (.skl), please refer to the SKL Format page.
 
Version Information

Distribution Version: 0.1, 07/13/2011
This program is tested on Windows XP, 2003 Server, Vista and Windows 7.  It is still not guaranteed to be bug-free and work properly with all versions of Windows. It is for education and research ONLY.

Related publications
 
1. Qi Shan, Brian Curless, and Tadayoshi Kohno, "Seeing through Obscure Glass", European Conference on Computer Vision (ECCV), 2010.

2. Ankit Gupta, Neel Joshi, Larry Zitnick, Michael Cohen, and Brian Curless, "Single Image Deblurring Using Motion Density Functions", European Conference on Computer Vision (ECCV), 2010.


Please contact Qi Shan for any question concerning the released program.
 
Examples

Input images Results Kernels
   


CSE logo Computer Science & Engineering
Box 352350, University of Washington
Seattle, WA  98195-2350
(206) 543-1695
[comments to Qi Shan]
Privacy policy and terms of use