Removing camera shake from a single image

Removing camera shake from a single image

Presented at SIGGRAPH 2006, Boston

Download PDF

Rob Fergus   Barun Singh   Aaron Hertzmann   Sam T. Roweis  William T. Freeman  

Computer Science and Artificial Intelligence Lab (CSAIL)

Massachusetts Institute of Technology

and

Department of Computer Science

University of Toronto

Abstract

Camera shake during exposure leads to objectionable image blur and ruins many photographs.  Conventional blind deconvolution methods typically assume  frequency-domain constraints on images, or overly simplified parametric forms for the motion path during camera shake. Real camera motions can follow convoluted paths, and a spatial domain prior can better maintain visually salient image characteristics. We introduce a method to remove the effects of camera shake from seriously blurred images.  The method assumes a uniform camera blur over the image and negligible in-plane camera rotation. In order to estimate the blur from the camera shake, the user must specify an image region without saturation effects. We show results for a variety of digital photographs taken from personal photo collections.

Downloads

PDF version of our SIGGRAPH paper (10MB)

Download Adobe Acrobat Reader 7

New: PDF of slides for 1 hour presentation (28MB)

Shorted Powerpoint slides from SIGGRAPH presentation (59MB)

Videos from the presentation:

Slow motion, pt.1 (wmv, 4.3MB), Slow motion, pt.2 (wmv, 10.5MB)

Camera Trajectory (avi, 4.9MB), Normal speed (wmv, 4.6MB)

** A zip file including ppt and all the videos (78MB)

Some pictures from the paper (11.8MB)
Includes inferred blur kernels.
N.B. Please take care not to compress the images - compression artifacts look very much like blur.
I will add more images to the ZIP file as I get permission from their owners.
If you are from the media, please email me before using them.

Download Matlab source code: zip
Here is the accompanying README file.

Acknowledgements
We'd like to thank James Miskin and David Mackay for putting their source code on the web.

We'd like to thank: Antonio Torralba (the camera guy!), Fredo Durand, Berthold Horn, Don Geman for their comments and suggestions.

We would like the thank the following people for supplying us with blurred images for the paper and presentation: Lyndsey Pickup, Omar Khan, Reinhard Klette, Michael Lewicki, Pietro Perona, Mukta Prasad, Ian Reid and Elizabeth Van Ruitenbeek.

Last update: Aug 3, 2006. Thanks for Ce Liu for the page format.