Comparisons
 
Navigation:    Main index    Single lens results    Standard camera lens and multispectral camera results
 
Move your mouse over captions to switch between blurred input and deconvolved result.
 

 
Comparison to independently deconvolved channels:
To show the effect of using our cross-channel prior we compare here the result with individually deconvolved channels (using [LEVIN 2007]) with our result using the cross-channel prior. Below "Original" is the original blurred image, "Individual" is the image with individually deconvolved channels, "LevinCross" is the result of extending Levin's IRLS method with our prior and "Ours" is our result using the cross-channel prior.
 
Original     Individual     LevinCross     Ours
 
Original     Individual     Ours
Further scene:

Meter image: Original, Individual, Ours
 

 
Comparison to [Schuler et al. 2011] on data with low chromatic aberration:

Below we first compare here our method to that of [Schuler et al. 2010] using their data with low chromatic aberration. The original image, three results with different regularizations, and the result of [Schuler et al. 2010] is shown below in the full image and in a patch. Note that the deconvolved patches might be slightly misaligned since we got PSFs from the authors that were not correctly centered. The image quality (especially sharpness) can however be evaluated without a perfect alignment.
 
Original     HighRegularization     MediumRegularization     LowRegularization     ReferenceSchuler
 
Original     HighRegularization     MediumRegularization     LowRegularization     ReferenceSchuler
Further scene:

Window image: Original, OurResult, ReferenceSchuler
 

 
Comparison to [Schuler et al. 2011] with high chromatic aberrations:

Now we compare our method to that of [Schuler et al. 2010] for datasets with high chromatic aberrations. The blurred images are the ones shown in the single lens section of the supplemental material. Below we show our deconvolution result and the deconvolution result using Schuler et al.'s method with their YUV prior. Note that for the large chromatic aberrations in the images below their method completely breaks down.
 
Our reconstruction     Reconstruction with YUV-prior
 
Our reconstruction     Reconstruction with YUV-prior
 
Our reconstruction     Reconstruction with YUV-prior
 
 
 
 
References
 
LEVIN, A., FERGUS, R., DURAND, F., AND FREEMAN, W. T., 2007. Deconvolution using natural image priors.

SCHULER, C. J., HIRSCH, M., HARMELING, S., AND SCHOLKOPF, B. 2011. Non-stationary correction of optical aberrations. In ICCV, IEEE.