ALGORITHM FOR MINIMIZING LATENT SHARP IMAGE COST FUNCTION AND POINT SPREAD FUNCTION COST FUNCTION WITH A SPATIAL MASK IN A REGULARIZATION TERM

• US 20140355901A1
• Filed: 03/11/2013
• Published: 12/04/2014
• Est. Priority Date: 03/29/2012
• Status: Active Grant

• Current Assignee: Nikon Corporation
• Original Assignee: Nikon Corporation
• Inventors: Tezaur, Radka
• Granted Patent: US 9,262,815 B2
• Time in Patent Office: Days
• US Class Current: 382/255
• CPC Class Codes: G06T 2207/10004 Still image; Photographic image
• CPC Class Codes: G06T 2207/20048 Transform domain processing
• CPC Class Codes: G06T 2207/20201 Motion blur correction
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening
• CPC Class Codes: G06T 5/75 Unsharp masking

Abstract

A method for deblurring a blurry image (18) includes utilizing a spatial mask and a variable splitting technique in the latent sharp image estimation cost function. Additionally or alternatively, the method can include the utilizing a spatial mask and a variable splitting technique in the PSF estimation cost function. The spatial mask can be in a regularization term of either or both the latent sharp image estimation cost function and the PSF cost function. The latent sharp image estimation cost function can be used for non-blind deconvolution. Alternatively, one or both cost functions can be used for blind deconvolution.

Claims

1. A method for deblurring a blurry image, the method comprising the steps of:

utilizing a first spatial mask in a fidelity term and a second spatial mask in a regularization term of a latent sharp image estimation cost function used for deconvolution of the blurry image; and

utilizing a variable splitting technique in the latent sharp image function, the variable splitting technique including introducing a first auxiliary variable and a first penalty term into the latent sharp image estimation cost function to split the operations of convolution and element-wise multiplication.

2. The method of claim 1 wherein the latent sharp image estimation cost function is minimized over an unknown latent sharp image L and the first auxiliary variable in an alternating fashion.

3. The method of claim 2 further comprising the step of minimizing the latent sharp image estimation cost function using fast Fourier transform.

4. The method of claim 1 further comprising the step of minimizing the latent sharp image estimation cost function using fast Fourier transform.

5. The method of claim 1 wherein the step of utilizing a variable splitting technique includes introducing a second auxiliary variable into the latent sharp image estimation cost function.

6. The method of claim 5 wherein the step of utilizing a variable splitting technique includes introducing a second penalty term into the latent sharp image estimation cost function.

7. The method of claim 1 further comprising the steps of (i) utilizing a spatial mask in a regularization term of a PSF cost function used for a PSF kernel estimation of the blurry image, and (ii) utilizing a variable splitting technique in the PSF cost function, the including introducing a second auxiliary variable and a second penalty term into the PSF estimation cost function.

8. The method of claim 7 wherein the PSF estimation cost function is minimized over an unknown PSF kernel K and the second auxiliary variable in an alternating fashion.

9. The method of claim 8 further comprising the step of minimizing the PSF cost function using fast Fourier transform.

10. A method for deblurring a blurry image, the method comprising the steps of:

utilizing a first spatial mask in a regularization term of a PSF cost function used to estimate a PSF blur kernel of the blurry image; and

utilizing a variable splitting technique in the PSF cost function, the variable splitting technique including introducing a first auxiliary variable and a first penalty term into the PSF estimation cost function to split the operations of convolution and element-wise multiplication.

11. The method of claim 10 wherein the PSF estimation cost function is minimized over an unknown PSF blur kernel and the first auxiliary variable in an alternating fashion.

12. The method of claim 11 further comprising the step of minimizing the PSF estimation cost function using fast Fourier transform.

13. The method of claim 10 further comprising the step of minimizing the PSF estimation cost function using fast Fourier transform.

14. The method of claim 10 wherein the PSF estimation cost function includes a fidelity term in addition to the regularization term.

15. The method of claim 10 wherein the PSF estimation cost function is a modified regularized least squares cost function that includes one or more fidelity terms and one or more regularization terms.

16. The method of claim 10 further comprising the step of utilizing a second spatial mask in the fidelity term of the PSF cost function.

17. The method of claim 16 wherein the step of utilizing a variable splitting technique in the PSF cost function includes introducing a second auxiliary variable and a second penalty term into the PSF estimation cost function.