Method conducting smoothing processing in a linear direction having a minimized magnitude of density change

• US 5,561,724 A
• Filed: 06/10/1994
• Issued: 10/01/1996
• Est. Priority Date: 06/11/1993
• Status: Expired due to Term

• Current Assignee: Hitachi Medical Corporation (Hitachi, Ltd.)
• Original Assignee: Hitachi Medical Corporation (Hitachi, Ltd.)
• Inventors: Sano, Koichi, Kido, Kunihiko, Taguchi, Jun'ichi
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Johns, Andrew W.
• Application Number:
  US08/258,382
• Time in Patent Office: 844 Days
• Field of Search: 382/254, 382/260, 382/264, 382/266, 382/275, 358/447, 358/463, 364/724.01, 348/607
• US Class Current: 382/264
• CPC Class Codes: G06T 5/20 using local operators

Abstract

A noise filtering method including the steps of calculating a magnitude E_i of density change for each of a plurality of linear direction components i in a local region around each of pixels of an inputted two-dimensional image, detecting such a direction as to minimize the magnitude E_i of density change, conducting smoothing processing by deriving an arithmetic mean of densities of pixels in the local region arranged in the detected direction, and thereby reducing noise of the inputted image.

Claims

1. A noise filtering method implemented using a computer for enhancing images, said method comprising the steps of:

calculating a magnitude E_i of density change for each of a plurality of linear direction components i in a local region around each pixel of an input image; and

conducting one-dimensional smoothing processing for each pixel of the input image in a linear direction of one of said plurality of linear direction components, said linear direction for each pixel being selected on a basis of having a minimum said magnitude E_i of density change for that pixel as calculated in said calculating step.

2. A noise filtering method according to claim 1, wherein a sum of absolute values of density differences between adjacent pixels located on a line of each linear direction component i is calculated and adopted as the magnitude E_i of density change.

3. A noise filtering method according to claim 1, wherein the magnitude E_i of density change is derived from a pixel value included in a 3*3 matrix.

4. A noise filtering method according to claim 3, wherein said plurality of linear direction components i are linear direction components of four directions (i=1 to 4) forming an angle of 45 degrees with each other.

5. A noise reducing filter according to claim 1, wherein the magnitude E_i of density change is derived from a pixel value included in a 5*5 matrix.

6. A noise filtering method according to claim 5, wherein said plurality of linear direction components i are linear direction components of eight directions (i=1 to 8) forming an angle with each other.

7. A noise filtering method according to claim 1, further comprising the step of determining a matrix size for smoothing, smoothing processing being conducted by using densities of a plurality of pixels along said linear direction of the input image in the determined matrix size for smoothing.

8. A noise filtering method according to claim 7, wherein with respect to a minimum value E_min of the magnitude E_i of density change calculated for each linear direction component i, a matrix size for smoothing is made larger as E_min becomes smaller.

9. A noise filtering method implemented using a computer for enhancing images, said method comprising the steps of:

calculating a magnitude E_i of density change for each of a plurality of linear direction components i in a local region around each pixel of an input image;

detecting a minimum value E_min of magnitude of density change for said each pixel, determining, on the basis of said minimum value E_min, whether two-dimensional smoothing should be preformed or one-dimensional smoothing should be performed, and detecting such a direction as to minimize E_i as a direction for smoothing in case one-dimensional smoothing is to be performed; and

performing plane smoothing processing in case the determined smoothing method is two-dimensional smoothing mid performing one-dimensional smoothing processing in the detected direction in case the determined smoothing method is one-dimensional smoothing.

10. A noise filtering method according to claim 9, wherein when said minimum value E_min of the magnitude E_i of density change is equal to or less than a predetermined threshold value, two-dimensional smoothing is performed with respect to the pixel.

11. A noise filtering method implemented using a computer for enhancing images, said method comprising the steps of:

calculating a magnitude E_i of density change for each of a plurality of linear direction components i in a local region around each pixel of an input image;

detecting such a direction as to minimize said magnitude E_i of density change;

expanding a smoothing region in the direction minimizing the magnitude E_i of density change, from an end of said local region in said detected direction toward an outside of said local region; and

conducting smoothing processing in the expanded region for each pixel.

12. A noise filtering method implemented using a computer for enhancing images, said method comprising the steps of:

successively determining, for each pixel of an input image, whether directionality of a local structure is present in each matrix size while monotonically decreasing the matrix size, and deciding a maximum matrix size in which directionality of the local structure can be judged to be present;

detecting, for said each pixel, a direction of linear direction component as to minimize the magnitude E_i of density change in the decided matrix size; and

conducting smoothing processing in the direction detected according to the matrix size decided for each pixel.

13. A noise filtering method according to claim 12, wherein it is determined whether directionality of a local structure is present in each of matrix sizes 5*5, 3*3 and 1*1.

14. A noise reducing filter according to claim 12, wherein it is determined whether directionality of a local structure is present in each of matrix sizes 3*3 and 1*1.

15. A noise filtering method according to claim 12, comprising the steps of:

calculating the magnitude E_i of density change of each linear direction component (i=1 to m) in a region of matrix size n*n for said each pixel;

deriving a maximum value E_max and a minimum value E_min of the magnitude E_i (i=1 to m) of density change for each pixel;

judging, for each pixel, directionality to be present in a local structure in a region having a matrix size of n*n of the pixel, when the value of D=E_max -E_min is larger than a predetermined threshold value T; and

judging, for each pixel, directionality to be not present in a local structure in a region having a matrix size of n*n of the pixel, when the value of D=E_max -E_min is smaller than the predetermined threshold value T.

16. A noise filtering method according to claim 12, comprising the steps of:

calculating the magnitude E_i of density change of each linear direction component (i=1 to m) in a region of matrix size n*n for said each pixel;

deriving a maximum value E_max and a minimum value E_min of the magnitude E_i (i=1 to m) of density change for each pixel;

judging, for each pixel, directionality to be present in a local structure in a region having a matrix size of n*n of the pixel, when the relation E_min <

p * E_max is satisfied for a predetermined threshold value p (0<

p≦

1); and

judging, for each pixel, directionality to be not present in a local structure in a region having a matrix size of n*n of the pixel, when the relation E_min ≧

p * E_max is satisfied.

17. A noise filtering method according to claim 12, comprising the steps of:

calculating the magnitude E_i of density change of each linear direction component (i=1 to m) in a region of matrix size n*n for said each pixel;

deriving a minimum value E_min of the magnitude E_i (i=1 to m) of density change for each pixel;

judging directionality to be present in a local structure in a region having a matrix size of n*n of the pixel, when the relation E_min <

T is satisfied for a predetermined threshold value T; and

judging directionality to be not present in a local structure in a region having a matrix size of n*n of the pixel, when the relation E_min ≧

T is satisfied.

18. A noise filtering method implemented using a computer for enhancing images, said method comprising the steps of:

calculating a magnitude E_i of density change for each of a plurality of linear direction components i in a local region around each pixel of an input image;

detecting a minimum value E_min of magnitude of density change for said each pixel, determining, on the basis of said minimum value E_min, whether two-dimensional smoothing should be preformed or one-dimensional smoothing should be performed, and detecting a linear direction of one of said plurality of linear direction components for smoothing in case one-dimensional smoothing is to be performed, said linear direction being detected on a basis of having a minimum said magnitude E_i of density change as calculated in said calculating step; and

performing plane smoothing processing in case the determined smoothing method is two-dimensional smoothing and performing one-dimensional smoothing processing in the detected direction in case the determined smoothing method is one-dimensional smoothing.

19. A noise filtering method implemented using a computer for enhancing images, said method comprising the steps of:

calculating a magnitude E_i of density change for each of a plurality of linear direction components i in a local region around each pixel of an input image;

detecting a linear direction of one of said plurality of linear direction components so as to minimize said magnitude E_i of density change as calculated in said calculating step;

expanding a smoothing region in said linear direction minimizing the magnitude E_i of density change from said detecting step, from an end of said local region in said linear direction toward an outside of said local region; and

conducting one-dimensional smoothing processing in the expanded region for each pixel.

20. A noise filtering method implemented using a computer for enhancing images, said method comprising the steps of:

successively determining, for each pixel of an input image, whether directionality of a local structure is present in each matrix size while monotonically decreasing the matrix size, and deciding a maximum matrix size in which directionality of the local structure can be judged to be present;

detecting, for said each pixel, a linear direction of one of a plurality of linear direction components i in a local region around each pixel, said linear direction being detected on a basis of having a minimum magnitude E_i of density change of all said plurality of linear direction components in a decided matrix size from said determining step; and

conducting one-dimensional smoothing processing in said linear direction detected according to the matrix size decided for each pixel.