Systems and methods for deblurring data corrupted by shift variant blurring

• US 7,860,333 B2
• Filed: 01/09/2007
• Issued: 12/28/2010
• Est. Priority Date: 01/09/2007
• Status: Expired due to Fees

• Current Assignee: University of Utah Research Foundation (State of Utah)
• Original Assignee: University of Utah Research Foundation (State of Utah)
• Inventors: Zeng, Gengsheng Lawrence
• Primary Examiner(s): PATEL, KANJIBHAI B
• Application Number:
  US11/621,555
• Publication Number:
  US 20080166063A1
• Time in Patent Office: 1,449 Days
• Field of Search: 382/128, 382/205, 382/254, 382/255, 382/260, 382/275, 382/280, 382/284, 382/285, 382/312, 358/3.26
• US Class Current: 382/260
• CPC Class Codes: G06T 2207/10104 Positron emission tomography [PET]
• CPC Class Codes: G06T 2207/10108 Single photon emission computed tomography [SPECT]
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening

Abstract

The present invention provides methods, systems and machine readable medium including machine readable code for deblurring data corrupted by shift variant blurring. A first version of data having shift variant blurring characterized by a first shift variant point spread function is provided. A target shift invariant point spread function is selected. A second shift variant point spread function is derived wherein a combination of the first and second shift variant point spread functions generates the target shift invariant point spread function. The second shift variant point spread function is applied to the first version of the data thereby generating a second version of the data having shift invariant blurring characterized by the target shift invariant point spread function. A linear shift invariant filter is applied to the second version of the data thereby generating a deblurred version of the data.

Claims

1. A method of deblurring data corrupted by shift variant blurring, the method comprising:

providing a first version of data having shift variant blurring characterized by a first shift variant point spread function;

selecting a target shift invariant point spread function;

deriving a second shift variant point spread function wherein a combination of the first and second shift variant point spread functions generates the target shift invariant point spread function;

applying the second shift variant point spread function to the first version of the data thereby generating a second version of the data having shift invariant blurring characterized by the target shift invariant point spread function; and

applying a linear shift invariant filter to the second version of the data thereby generating a deblurred version of the data.

2. The method of claim 1, wherein providing a first version of data comprises providing a first version of data selected from a group consisting of one dimensional data, two-dimensional data, and three dimensional data.

3. The method of claim 1, further comprising:

providing first and second two dimensional images of a target taken from first and second viewing angles, respectively; and

reconstructing the first version of the data from the first and second two dimensional images of the target, the reconstructed first version of the data being a three dimensional image.

4. The method of claim 3, wherein providing first and second two dimensional images of a target comprises providing first and second two dimensional images of a target as captured by a nuclear imaging assembly.

5. The method of claim 3, wherein reconstructing the first version of the data from the first and second two dimensional images of the target comprises reconstructing the first version of the data from the first and second two dimensional images of the target using a reconstruction algorithm selected from a group consisting of a filtered backprojection (FBP) reconstruction algorithm and an iterative reconstruction algorithm.

6. The method of claim 3, wherein reconstructing the first version of the data from the first and second two dimensional images of the target comprises reconstructing the first version of the data from the first and second two dimensional images of the target using Novikov'"'"'s filtered backprojection (FBP) reconstruction algorithm.

7. The method of claim 1, wherein the first version of data comprises a three dimensional image, the method further comprising:

determining a collimator specific point spread function characterizing a first component of the shift variant blurring attributable to a collimator specific blurring effect in an image domain;

determining a target specific scattering point spread function characterizing a second component of the shift variant blurring attributable to a target specific scattering effect in an image domain; and

deriving the first variant point spread function based on the collimator specific point spread function and the target specific point spread function.

8. The method of claim 7, wherein determining a collimator specific point spread function in the image domain comprises determining the collimator specific point spread function based on collimator data selected from a group consisting of manufacturer supplied collimator specification data, collimator projection data, and collimator detector point spread function data.

9. The method of claim 7, wherein determining a collimator specific point spread function in the image domain comprises determining a collimator specific point spread function using a reconstruction algorithm selected from a group consisting of a filtered backprojection (FBP) reconstruction algorithm and an iterative reconstruction algorithm.

10. The method of claim 1, wherein selecting a target shift invariant point spread function comprises selecting a target shift invariant point spread function from a group consisting of a point spread function of a blurriest section of the first version of the data, a point spread function that exceeds a point spread function of blurriest section of the first version of the data, and a point spread function of a relatively blurry section of the first version of the data.

11. The method of claim 1, wherein deriving a second shift variant point spread function comprises:

deteimining a Fourier transform of the target shift invariant point spread function;

determining a Fourier transform of the first shift variant point spread function;

generating a Fourier transform of the second shift variant point spread function by dividing the Fourier transform of the target shift invariant point spread function by the Fourier transform of the first shift variant point spread function; and

determining the inverse Fourier transform of the second shift variant point spread function thereby generating the second shift variant point spread function.

12. The method of claim 1, wherein deriving a second shift variant point spread function comprises:

determining a variance of the target shift invariant point spread function;

determining a variance of the first shift variant point spread function;

generating a variance of the second shift variant point spread function by subtracting the variance of the first shift variant pint spread function from the variance of the target shift invariant point spread function; and

deriving the second shift variant point spread function from the variance of the second shift variant point spread function.

13. The method of claim 1, wherein applying the second shift variant point spread function to the first version of the data thereby generating a second version of the data having shift invariant blurring characterized by the target shift invariant point spread function comprises applying the second shift variant point spread function to the first version of the data using a rotational convolution method.

14. The method of claim 1, wherein applying a linear shift invariant filter to the second version of the data thereby generating a deblurred version of the data comprises applying a linear shift invariant filter selected from a group consisting of a non-iterative filter, an iterative filter, a convolution procedure based filter, a frequency domain multiplication procedure based filter, and a fast Fourier domain de-blurring filter.

15. A non-transitory machine readable medium for storing a machine executable program for deblurring data corrupted by shift variant blurring comprising:

machine readable code for providing a first version of data having shift variant blurring characterized by a first shift variant point spread function;

machine readable code for selecting a target shift invariant point spread function;

machine readable code for deriving a second shift variant point spread function wherein a combination of the first and second shift variant point spread functions generates the target shift invariant point spread function;

machine readable code for applying the second shift variant point spread function to the first version of the data thereby generating a second version of the data having shift invariant blurring characterized by the target shift invariant point spread function; and

machine readable code for applying a linear shift invariant filter to the second version of the data thereby generating a deblurred version of the data.

16. The non-transitory machine readable medium of claim 15, wherein the machine readable code for providing a first version of data comprises machine readable code for providing a first version of data selected from a group consisting of one dimensional data, two dimensional data, and three dimensional data.

17. The non-transitory machine readable medium of claim 15, further comprising:

machine readable code for reconstructing the first version of the data from the first and second different angular views of the target, the reconstructed first version of the data being a three dimensional image.

18. The non-transitory machine readable medium of claim 17, wherein the machine readable code for providing first and second different angular views of a target comprises machine readable code for retrieving the first and second different angular views of the target from a nuclear imaging assembly.

19. The non-transitory machine readable medium of claim 17, wherein the machine readable code for reconstructing the first version of the data from the first and second different angular views of the target comprises machine readable code for reconstructing the first version of the data from the first and second different angular views of the target using a filtered backprojection (FBP) reconstruction algorithm.

20. The non-transitory machine readable medium of claim 17, wherein the machine readable code for reconstructing the first version of the data from the first and second different angular views of the target comprises machine readable code for reconstructing the first version of the data from the first and second different angular views of the target using Novikov'"'"'s filtered backprojection (FBP) reconstruction algorithm.

21. The non-transitory machine readable medium of claim 15, wherein the first version of data comprises a three dimensional image, the machine readable medium further comprising:

machine readable code for determining a collimator specific point spread function characterizing a first component of the shift variant blurring attributable to a collimator specific blurring effect in an image domain;

machine readable code for determining a target specific scattering point spread function characterizing a second component of the shift variant blurring attributable to a target specific scattering effect in an image domain; and

machine readable code for deriving the first variant point spread function based on the collimator specific point spread function and the target specific point spread function.

22. The non-transitory machine readable medium of claim 21, wherein the machine readable code for determining a collimator specific point spread function in the image domain comprises machine readable code for determining the collimator specific point spread function based on collimator data selected from a group consisting of manufacturer supplied collimator specification data, collimator projection data, and collimator detector point spread function data.

23. The non-transitory machine readable medium of claim 21, wherein the machine readable code for determining a collimator specific point spread function in the image domain comprises machine readable code for determining a collimator specific point spread function using Novikov'"'"'s filtered backprojection (FBP) algorithm.

24. The non-transitory machine readable medium of claim 15, wherein the machine readable code for selecting a target shift invariant point spread function comprises machine readable code for selecting a target shift invariant point spread function from a group consisting of a point spread function of a blurriest section of the first version of the data, a point spread function that exceeds a point spread function of a blurriest section of the first version of the data, and a point spread function of a relatively blurry section of the first version of the data.

25. The non-transitory machine readable medium of claim 15, wherein the machine readable code for deriving a second shift variant point spread function comprises:

machine readable code for determining a Fourier transform of the target shift invariant point spread function;

machine readable code for determining Fourier transform of the first shift variant point spread function;

machine readable code for generating a Fourier transform of the second shift variant point spread function by dividing the Fourier transform of the target shift invariant point spread function by the Fourier transform of the first shift variant point spread function; and

machine readable code for determining the inverse Fourier transform of the second shift variant point spread function thereby generating the second shift variant point spread function.

26. The non-transitory machine readable medium of claim 15, wherein the machine readable code for deriving a second shift variant point spread function comprises:

machine readable code for determining a variance of the target shift invariant point spread function;

machine readable code for determining a variance of the first shift variant point spread function;

machine readable code for generating a variance of the second shift variant point spread function by subtracting the variance of the first shift variant point spread function from the variance of the target shift invariant point spread function; and

machine readable code for deriving the second shift variant point spread function from the variance of the second shift variant point spread function.

27. The non-transitory machine readable medium of claim 15, wherein the machine readable code for applying the second shift variant point spread function to the first version of the data thereby generating a second version of the data having shift invariant blurring characterized by the target shift invariant point spread function comprises machine readable code for applying the second shift variant point spread function to the first version of the data using a rotational convolution method.

28. The non-transitory machine readable medium of claim 15, wherein the machine readable code for applying a linear shift invariant filter to the second version of the data thereby generating a deblurred version of the data comprises machine readable code for applying a linear shift invariant filter selected from a group consisting of a non-iterative filter, an iterative filter, a convolution procedure based filter, a frequency domain multiplication procedure based filter, and a fast Fourier domain de-blurring filter.

29. A method of deblurring data corrupted by shift variant blurring, the method comprising:

providing a first version of data having shift variant blurring;

blurring the first version of the data further using a shift variant blurring kernel thereby generating a second version of data having shift invariant blurring; and

applying a linear shift invariant filter to the second version of the data thereby generating a deblurred version of the data.