Automatically determining the size of a blur kernel

1. A computer-implemented method, comprising:

accessing an image having at least one blurred region;

automatically, without user input, determining a first value for a first size for a blur kernel for the at least one blurred region;

automatically, without user input, determining a second value for a second size for the blur kernel for the at least one blurred region;

determining a suggested size for the blur kernel based on the first value and the second value; and

deblurring the accessed image by deconvolving the at least one blurred region with the blur kernel of the suggested size.

2. The method of claim 1 , wherein different mathematical algorithms are used to determine the first and second values.

3. The method of claim 1 , wherein the first value for the first size of the blur kernel is determined by an autocorrelation-based size estimator, and the second value for the second size of the blur kernel is determined by a latent image-based size estimator.

4. The method of claim 1 , further comprising:

generating a derivative image from the accessed image;

autocorrelating the derivative image and the accessed image to obtain an autocorrelation result; and

determining the first size for the blur kernel from the autocorrelation result.

5. The method of claim 4 , wherein the determining of the first size is derived from nonzero elements of the autocorrelation result.

6. The method of claim 5 , wherein the autocorrelation result is filtered to remove nonzero elements less than a threshold value.

7. The method of claim 1 , wherein determining the first value for the first size for the blur kernel includes:

calculating a derivative image for the accessed image;

determining an autocorrelation map for the derivative image; and

identifying a connected component of nonzero elements displayed in the autocorrelation map, wherein determining the second value for the second size of the blur kernel includes:

determining a latent image for the accessed image;

estimating a blur kernel from the latent image; and

identifying a connected component of nonzero elements displayed in the estimated blur kernel.

8. The method of claim 1 , further comprising:

applying a shock filter to the accessed image to restore sharp edges to provide filtered results;

computing one or more gradient maps from the filtered results; and

estimating the second size based on the gradient maps.

9. The method of claim 8 , wherein gradient values below a gradient threshold are removed from the one or more gradient maps.

10. The method of claim 1 , wherein determining the suggested size for the blur kernel based on the first value and the second value includes:

calculating an average value of the first and second values; and

selecting the average value as the suggested size for the blur kernel.

11. The method of claim 1 , further comprising:

receiving input identifying a user-selected size for the blur kernel; and

determining the suggested size for the blur kernel based on the first value, the second value and the user-selected size.

12. An image deblurring system, comprising:

a processor;

an image module, coupled with the processor, the image module operated on the processor, to access an image having at least one blurred region; and

a blur kernel sizing module, coupled with the processor, the blur kernel sizing module to:

automatically, without user input, determine a first value for a first size of a blur kernel for the at least one blurred region;

automatically, without user input, determine a second value for a second size of the blur kernel for the at least one blurred region;

determine a suggested size for the blur kernel based on the first value and the second value; and

deblurring the accessed image by deconvolving the at least one blurred region with the blur kernel of the suggested size.

13. The system of claim 12 , wherein different mathematical algorithms are used to determine the first and second values.

14. The system of claim 12 , wherein the first value for the first size of the blur kernel is determined by an autocorrelation-based size estimator, and the second value for the second size of the blur kernel is determined by a latent image-based size estimator.

15. The system of claim 12 , wherein determining the suggested size for the blur kernel based on the first value and the second value includes:

calculating an average value of the first and second values; and

selecting the average value as the suggested size for the blur kernel.

16. The system of claim 12 , wherein determining the first value for the first size for the blur kernel includes:

calculating a derivative image for the accessed image;

determining an autocorrelation map for the derivative image; and

identifying a connected component of nonzero elements displayed in the autocorrelation map, wherein determining the second value for the second size of the blur kernel includes:

determining a latent image for the accessed image;

estimating a blur kernel from the latent image; and

identifying a connected component of nonzero elements displayed in the estimated blur kernel.

17. A non-transitory computer-readable storage medium including instructions, when executed by a computer, cause the computer to perform operations

comprising:

accessing an image having at least one blurred region;

automatically, without user input, determining a first value for a first size for a blur kernel;

automatically, without user input, determining a second value for a second size for the blur kernel;

determining a suggested size for the blur kernel based on the first value and the second value; and

deblurring the accessed image by deconvolving the at least one blurred region with the blur kernel of the suggested size.

18. The non-transitory computer-readable storage medium of claim 17 , wherein determining the first value comprises determining the first value for the first size of the blur kernel based on an autocorrelation-based estimation, and determining the second value comprises determining the second value for the second size of the blur kernel based on a latent image-based estimation.