Methods for Reducing Row and Column Patterns in a Digital Image

• US 20140320496A1
• Filed: 03/31/2014
• Published: 10/30/2014
• Est. Priority Date: 05/19/2011
• Status: Active Grant

• Current Assignee: Foveon, Inc. (Avis Budget Group Incorporated)
• Original Assignee: Foveon, Inc. (Avis Budget Group Incorporated)
• Inventors: Yasan, Alireza, Rastislav, Lukac
• Granted Patent: US 9,240,035 B2
• Time in Patent Office: Days
• US Class Current: 345/428
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06V 10/30 Noise filtering

Abstract

A method for performing column/row pattern suppression in a digital input image includes creating a smoothed version of the input image by averaging a set of columns/rows neighboring around the column/row being corrected. A difference image is constructed by subtracting the smoothed image from the input image. New column/row intensities are computed from the difference image. An output image is constructed with suppressed column/row patterns by subtracting the new column/row intensities from the input image.

Claims

1. A method for performing column/row pattern suppression in a digital input image, comprising:

a) dividing the input image into regions;

b) selecting a region that has not been processed;

c) generating a smoothed version of the input image in the selected region by averaging a set of columns/rows neighboring around the column/row being corrected;

d) constructing a difference residual image in the selected region by subtracting the smoothed image from the input image in the selected region;

e) computing new column/row intensities in the selected region from the difference residual image;

f) constructing an output image in the selected region with suppressed column/row patterns by subtracting the new column/row intensities in the selected region from the input image in the selected region, including defining a maximum allowable correction amount and performing no correction on a column/row if the column/row average of the difference image exceeds the maximum allowable correction amount;

repeating b) through f) until all of the regions of the input image have been processed; and

combining all of the output images from all of the processed regions to generate an output image.

2. The method of claim 1, wherein:

dividing the input image into regions comprises dividing the input image into color channels; and

combining all of the output images from all of the processed regions to generate an output image comprises combining all of the output images from all of the processed color channels to generate an output image.

3. The method of claim 1, wherein:

dividing the input image into regions comprises dividing the input image into spatial regions; and

combining all of the output images from all of the processed regions to generate an output image comprises combining all of the output images from all of the processed spatial regions to generate an output image.

4. The method of claim 1, wherein:

dividing the input image into regions comprises dividing the input image into intensity bands; and

combining all of the output images from all of the processed regions to generate an output image comprises combining all of the output images from all of the processed intensity bands to generate an output image.

5. A method for performing column/row pattern suppression in a digital input image, comprising:

a) dividing the input image into spatial regions;

b) selecting a spatial region that has not been processed;

c) generating a smoothed version of the input image in the selected spatial region by averaging a set of columns/rows neighboring around the column/row being corrected;

d) constructing a difference residual image in the selected spatial region by subtracting the smoothed image from the input image in the selected spatial region;

e) computing new column/row intensities in the selected spatial region from the difference residual image;

f) constructing an output image in the selected spatial region with suppressed column/row patterns by subtracting the new column/row intensities in the selected spatial region from the input image in the selected spatial region, including defining a maximum allowable correction amount and performing no correction on a column/row if the column/row average of the difference image exceeds the maximum allowable correction amount;

repeating b) through f) until all of the spatial regions of the input image have been processed; and

combining all of the output images from all of the processed spatial regions to generate an output image.

6. The method of claim 5, wherein computing new column/row intensities from the difference image comprises ignoring pixels whose values change by more than a pass threshold.

7. The method of claim 5, wherein dividing the input image into spatial regions comprises dividing the input image into non-overlapping spatial regions.

8. The method of claim 5, wherein dividing the input image into spatial regions comprises dividing the input image into overlapping spatial regions.

9. The method of claim 5, wherein computing new column/row intensities in the selected spatial region from the difference residual image comprises subtracting a correction value from the difference residual image.

10. A method for performing column/row noise suppression in a digital input image, comprising:

a) dividing the input image into intensity bands;

b) selecting an intensity band that has not been processed;

c) generating a smoothed version of the input image in the selected intensity band by averaging a set of columns/rows neighboring around the column/row being corrected;

d) constructing a difference image in the selected intensity band by subtracting the smoothed image from the input image in the selected intensity band;

e) computing new column/row intensities in the selected intensity band from the difference residual image;

f) constructing an output image in the selected intensity band with suppressed column/row patterns by subtracting the new column/row intensities in the selected intensity band from the input image in the selected intensity band, including defining a maximum allowable correction amount and performing no correction on a column/row if the column/row average of the difference image exceeds the maximum allowable correction amount;

repeating b) through f) until all of the intensity bands of the input image have been processed; and

combining all of the output images from all of the processed spatial regions to generate an output image.

11. The method of claim 10, wherein computing new column/row intensities from the difference image comprises ignoring pixels whose values change by more than a pass threshold.

12. The method of claim 10, wherein dividing the input image into intensity bands comprises dividing the input image into non-overlapping intensity bands.

13. The method of claim 10, wherein dividing the input image into intensity bands comprises dividing the input image into overlapping intensity bands.

14. The method of claim 10, wherein computing new column/row intensities in the selected intensity band from the difference residual image comprises subtracting a correction value from the difference residual image.

15. A method for performing column/row noise suppression in a digital input image, comprising:

generating a smoothed version of the input image in the selected region by applying a low pass filter to the input image;

generating a grayscale residual image as a function of the input image and the smoothed image;

computing new column/row intensities in the selected region from the grayscale residual image;

constructing an output image with suppressed column/row patterns by subtracting the new column/row intensities from the input image, including defining a maximum allowable correction amount and performing no correction on a column/row if the column/row average of the difference image exceeds the maximum allowable correction amount; and

combining all of the output images from all of the processed regions to generate an output image.

16. The method of claim 15, wherein computing new column/row intensities from the grayscale image comprises ignoring pixels whose values change by more than a pass threshold.