http://www.soe.ucsc.edu/~xzhu/images/seal_bl_1inch_72dpi.gif

UNIVERSITY OF CALIFORNIA
SANTA CRUZ


 

 

 

 

Removing Atmospheric Turbulence

Xiang Zhu and Peyman Milanfar

 

 

Abstract

To correct geometric distortion and reduce space and time-varying blur, a new approach is proposed in this paper capable of restoring a single high-quality image from a given image sequence distorted by atmospheric turbulence. This approach reduces the space and time-varying deblurring problem to a shift invariant one. It first registers each frame to suppress geometric deformation through B-spline based non-rigid registration. Next, a temporal regression process is carried out to produce an image from the registered frames, which can be viewed as being convolved with a space invariant near-diffraction-limited blur. Finally, a blind deconvolution algorithm is implemented to deblur the fused image, generating a final output. Experiments using real data illustrate that this approach can effectively alleviate blur and distortions, recover details of the scene and significantly improve visual quality.

Related Papers

X. Zhu and P. Milanfar, "Removing Atmospheric Turbulence via Space-Invariant Deconvolution", Accepted to IEEE Trans. on Pattern Analysis and Machine Intelligence, DOI: 10.1109/TPAMI.2012.82.

X. Zhu and P. Milanfar, "Stabilizing and Deblurring Atmospheric Turbulence", International Conference on Computational Photography (ICCP), Pittsburgh, PA, April 2011. See related talk.

Restoration Framework

[MATLAB CODE]

 

 

diag_prop3.JPG

 

 

 

Experimental Results

 

 

 

1. Moon Surface (410x380x80). This video is taken from a ground-based telescope.

 

       

Observed video                                                                            Registered Video


One Observed Frame


Proposed Approach


Please place your mouse over the links to show the corresponding image on the left pane. You do not need to click the links. The default image shown is the result obtained from our method.

 

 

 

2. Water Tower (300x220x80). Top part of a water tower located above the ground, imaged at a (horizontal) distance of 2.4 kilometers.

 

       

Observed Video                                                                            Registered Video


One Observed Frame


Proposed Approach


 

 

 

3. Chimney (237x237x100). This video is captured through hot air exhausted by a building's vent.

 

       

Observed video                                                                            Registered Video


One Observed Frame


EFF (Hirsch et al. CVPR2010)


Proposed Approach

 

 

 

4. Building (237x237x100). This video is captured through hot air exhausted by a building's vent.

       

Observed video                                                                            Registered Video


One Observed Frame


EFF (Hirsch et al. CVPR2010)


Proposed Approach

 

 

 

Acknowledgements

 

We would like to thank Prof. Mikhail A. Vorontsov from the Intelligent Optics Lab of the University of Maryland for allowing us to use the video data Water Tower, and thank Mr. Faisal A. Salem from University of Michigan and Dr. Joseph M. Zawodny from NASA Langley Research Center for providing us with the video Moon Surface. We also thank Mr. M. Hirsch and Dr. S. Harmeling from Max Plank Institute for Biological Cybernetics for sharing with us the sequences Chimney and Building.