Methods and apparatus for noise reduction filtering of images

• US 7,835,562 B2
• Filed: 07/23/2004
• Issued: 11/16/2010
• Est. Priority Date: 07/23/2004
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Avinash, Gopal B.
• Primary Examiner(s): Le; Vu
• Assistant Examiner(s): Vanchy, Jr.; Michael
• Application Number:
  US10/897,764
• Publication Number:
  US 20060018526A1
• Time in Patent Office: 2,307 Days
• Field of Search: 382/131, 382/132, 382/260
• US Class Current: 382/132
• CPC Class Codes: G06T 11/005 Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
• CPC Class Codes: G06T 2207/10081 Computed x-ray tomography [CT]
• CPC Class Codes: G06T 2207/20012 Locally adaptive
• CPC Class Codes: G06T 2207/20192 Edge enhancement; Edge preservation
• CPC Class Codes: G06T 2207/20221 Image fusion; Image merging
• CPC Class Codes: G06T 2211/424 Iterative
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/50 using two or more images, e.g. averaging or subtraction
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening
• CPC Class Codes: G06T 5/92 based on global image properties

Abstract

A method for reducing noise in images includes utilizing an original image to generate a segmentation mask and a filtered image, and utilizing the filtered image, the original image, the segmentation mask and separate blending parameters for structural and nonstructural portions of the filtered image to generate a blended image. The method further includes contrast matching the blended image with the original image to produce an output image.

Claims

1. A method for reducing noise in images, said method comprising:

acquiring a single original image using an electron beam computed tomographic scanner;

generating original image data associated with an input of the single original image;

generating segmentation mask data from the single original image and a single threshold, the segmentation mask data representing a segmentation mask of structural portions and non-structural portions in the single original image;

generating filtered image data from the segmentation mask data by scaling intensity values and using a connectivity approach, the filtered image data representing a filtered image;

defining a first blending parameter for the structural portions in the filtered image and a second blending parameter for the non-structural portions in the filtered image, the first blending parameter different than the second blending parameter;

generating blended image data from the filtered image data, the original image data, the segmentation mask data, and the blending parameters for; and

contrast matching the blended image data with the original image data to produce output image data, the output image data representing a single processed image.

2. A method in accordance with claim 1 further comprising:

shrinking the single original image in accordance with a parameter and X to produce shrunken image data; and

expanding the segmentation mask and the filtered image in accordance with the parameter X to produce expanded segmentation mask data and expanded filtered image data,

wherein said generating segmentation mask data and said generating filtered image data comprises filtering the shrunken image data to generate the segmentation mask data and the filtered image data, and said generating blended image data comprises blending the expanded filtered image data with the original image data utilizing the expanded segmentation mask data and the blending parameters to generate the blended image data.

3. A method in accordance with claim 2 wherein generating filtered image data further comprises anisotropically filtering the structural portions of the shrunken image data and isotropically filtering the non-structural portions of the shrunken image data.

4. A method in accordance with claim 3 wherein generating segmentation mask image data comprises:

determining gradient values for pixels of the shrunken image data;

computing a histogram of gradient magnitudes; and

separating the structure portions from the non-structure portions using the gradient histogram.

5. A method in accordance with claim 3 wherein said isotropically filtering comprises boxcar smoothing.

6. A method in accordance with claim 3 wherein said anisotropically filtering comprises filtering a structure region to extract domain orientation information and iteratively filtering structures along a dominant direction in a given neighborhood.

7. A method in accordance with claim 1 further comprising setting a gradient threshold as the single threshold and using the gradient threshold to separate the structure portions from the non-structure portions.

8. A method in accordance with claim 7 further comprising utilizing a dominant orientation and an orthogonal orientation to make a consistency decision.

9. A method in accordance with claim 1 wherein the single original image is a scanning electron beam computed tomography image.

10. An apparatus for reducing noise in images, said apparatus comprising a computer configured to:

acquire a single original image using an electron beam computed tomographic scanner;

generate original image data associated with an input of the single original image;

generate segmentation mask data from the single original image and a single threshold, the segmentation mask data representing a segmentation mask of structural portions and non-structural portions of the single original image;

generating filtered image data from the segmentation mask data by scaling intensity values and using a connectivity approach, the filtered image data representing a filtered image;

define a first blending parameter for the structural portions in the filtered image and a second blending parameter for the non-structural portions in the filtered image, the first blending parameter different than the second blending parameter;

,

generate blended image data from the filtered image data, the original image data, the segmentation mask data, and the blending parameters; and

contrast match the blended image data with the original image data to produce output image data.

11. A computer in accordance with claim 10 further configured to:

shrink the single original image in accordance with a parameter X to produce shrunken image data; and

expand the segmentation mask and the filtered image in accordance with the parameter X to produce expanded segmentation mask data and expanded filtered image data,

wherein, to generate the segmentation mask data and the filtered image data, said computer is configured to filter the shrunken image data to generate the segmentation mask data and the filtered image data, and to blend the expanded filtered image data with the original image data utilizing the expanded segmentation mask data and the separate blending parameters to generate the blended image data.

12. A computer in accordance with claim 11 wherein said computer is further configured to anisotropically filter the structural portions of the shrunken image data and isotropically filter the non-structural portions of the shrunken image data.

13. A computer in accordance with claim 12 wherein said computer is configured to:

determine gradient values for pixels of the shrunken image data;

compute a histogram of gradient magnitudes; and

separate the structure portions from the non-structure portions using the gradient histogram.

14. A computer in accordance with claim 13 further configured to set a gradient threshold as the single threshold and use the gradient threshold to separate the structure portions from the non-structure portions.

15. A computer in accordance with claim 14 further configured to utilize a dominant orientation and an orthogonal orientation to make a consistency decision.

16. A computer in accordance with claim 12 wherein, to isotropically filter, said computer is configured to utilize boxcar filtering.

17. A computer in accordance with claim 12 wherein, to anisotropically filter, said computer is configured to filter the structural portions to extract domain orientation information and iteratively filter the structural portions along a dominant direction in a given neighborhood.

18. A computer in accordance with claim 10 wherein said computer is configured as a data processing system in the electron beam computed tomography scanner.

19. A non-transitory machine readable medium comprising instructions recorded thereon, said instructions configured to instruct a computer to:

generate original image data associated with an input of a single original image;

generate segmentation mask from the single original image and a single threshold, the segmentation mask data representing a segmentation mask of structural portions and non-structural portions of the single original image;

generate filtered image data from the segmentation mask data by scaling intensity values and using a connectivity approach, the filtered image data representing a filtered image;

define a first blending parameter for the structural portions in the filtered image and a second blending parameter for the non-structural portions in the filtered image, the first blending parameter different than the second blending parameter;

generate blended image data from the filtered image data, the original image data, the segmentation mask data, and blending parameters; and

contrast match the blended image data with the original image data to produce output image data.

20. A machine readable medium in accordance with claim 19 further comprising thereon instructions configured to instruct the computer to:

shrink the single original image in accordance with a parameter X; and

expand the segmentation mask data and the filtered image data in accordance with the parameter X to produce expanded segmentation mask data and expanded filtered image data,

wherein, to generate the segmentation mask data and the filtered image data, said instructions are further configured to instruct the computer to filter the shrunken image data to generate the segmentation mask data and the filtered image data, and to blend the expanded filtered image data with the original image data utilizing the expanded segmentation mask data and the blending parameters to generate the blended image data.

21. A machine readable medium in accordance with claim 20 further comprising thereon instructions configured to instruct the computer to anisotropically filter the structural portions of the shrunken image data and isotropically filter the non-structural portions of the shrunken image data.

22. A machine readable medium in accordance with claim 21 further comprising thereon instructions configured to instruct the computer to:

determine gradient values for pixels of the shrunken image data;

compute a histogram of gradient magnitudes; and

separate the structure portions from the non-structure portions using the gradient histogram.

23. A machine readable medium in accordance with claim 22 further comprising thereon instructions configured to instruct the computer to set a gradient threshold as the single threshold and use the initial gradient threshold to the separate structure portions from the non-structure portions.

24. A machine readable medium in accordance with claim 23 further comprising thereon instructions configured to instruct the computer to utilize a dominant orientation and an orthogonal orientation to make a consistency decision.

25. A machine readable medium in accordance with claim 21 further comprising thereon instructions configured to instruct the computer to isotropically filter via boxcar filtering.

26. A machine readable medium in accordance with claim 21 further comprising thereon instructions configured to instruct the computer to filter the structural portions to extract domain orientation information and iteratively filter the structural portions along a dominant direction in a given neighborhood.

27. An electron beam computed tomographic (EBCT) scanner comprising:

a vacuum chamber;

an electron gun configured to produce an electron beam that passes through said vacuum chamber;

an assembly of targets;

a focusing and scanning system configured to focus the electron beam so that it impinges upon at least one target in said assembly of targets and to produce an X-ray beam;

an assembly of detectors configured to detect the produced X-ray beam that pass through an object and to produce a digitized intensity distribution of the detected X-ray beam that is representative of a single original image of the object; and

a data processing system configured to receive and process the single original image of the object, said data processing system further configured to;

generate original image data associated with the single original image produced using said EBCT scanner;

generate segmentation mask data from the single original image and a single threshold, the segmentation mask data representing a segmentation mask of structural portions and non-structural portions of the single original image;

generate filtered image data from the segmentation mask data by scaling intensity values and using a connectivity approach, the filtered image data representing a filtered image;

define a first blending parameter for the structural portions in the filtered image and a second blending parameter for the non-structural portions in the filtered image, the first blending parameter different than the second blending parameter;

generate blended image data from the filtered image data, the original image data, the segmentation mask data, and blending parameters; and

contrast match the blended image data with the original image data to produce output image data.

28. A scanner in accordance with claim 27 wherein said data processing system is further configured to:

shrink the single original image in accordance with a parameter X to produce shrunken image data; and

expand the segmentation mask data and the filtered image data in accordance with the parameter X to produce expanded segmentation mask data and expanded filtered image data,

wherein, to generate the segmentation mask data and the filtered image data, said data processing system is configured to filter the shrunken image data to generate the segmentation mask data and the filtered image data, and to blend the expanded filtered image data with the original image data utilizing the expanded segmentation mask data and the blending parameters to generate the blended image data.

29. A scanner in accordance with claim 28 wherein said data processing system is further configured to anisotropically filter the structural portions of the shrunken image data and isotropically filter the non-structural portions of the shrunken image data.

30. A scanner in accordance with claim 29 wherein said data processing system is configured to determine gradient values for pixels of the shrunken image data, compute a histogram of gradient magnitudes, and separate the structure portions from the non-structure portions using the gradient histogram.

31. A scanner in accordance with claim 30 wherein said data processing system is further configured to set a gradient threshold as the single threshold and use the gradient threshold to separate the structure portions from the non-structure portions.

32. A scanner in accordance with claim 31 wherein said data processing system is further configured to utilize a dominant orientation and an orthogonal orientation to make a consistency decision.

33. A scanner in accordance with claim 29 wherein said data processing system is further configured to isotropically filter via boxcar filtering.

34. A scanner in accordance with claim 29 wherein said data processing system is further configured to filter the structural portions to extract domain orientation information and to iteratively filter the structural portions along a dominant direction in a given neighborhood.