System and method for removing image noise

• US 9,563,938 B2
• Filed: 05/05/2014
• Issued: 02/07/2017
• Est. Priority Date: 11/19/2010
• Status: Active Grant

• Current Assignee: Analog Devices International Unlimited Company (Analog Devices, Inc.)
• Original Assignee: Analog Devices Global Unlimited Company (Analog Devices, Inc.)
• Inventors: Singh, Raka, Mahapatra, Rajesh, Malik, Gaurav
• Primary Examiner(s): Seth, Manav
• Application Number:
  US14/269,979
• Publication Number:
  US 20140241636A1
• Time in Patent Office: 1,009 Days
• US Class Current: 1/1
• CPC Class Codes: G06T 2207/20032 Median filtering
• CPC Class Codes: G06T 2207/20192 Edge enhancement; Edge preservation
• CPC Class Codes: G06T 5/30 Erosion or dilatation, e.g. thinning
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: G06V 10/30 Noise filtering
• CPC Class Codes: H04N 1/409 Edge or detail enhancement; Noise or error suppression
• CPC Class Codes: H04N 1/58 Edge or detail enhancement; Noise or error suppression, e.g. colour misregistration correction H04N1/62 takes precedence
• CPC Class Codes: H04N 9/646 for image enhancement, e.g. vertical detail restoration, cross-colour elimination, contour correction, chrominance trapping filters

Abstract

A system and method for removing noise from images are disclosed herein. An exemplary system includes an edge-detection-based adaptive filter that identifies edge pixels and non-edge pixels in an image and selects a filtering technique for at least one non-edge pixel based on a comparison of the at least one non-edge pixel to a neighboring pixel region, wherein such comparison indicates whether the at least one non-edge pixel is a result of low-light noise.

Claims

1. A system for removing noise from an image, the system comprising:

an edge-detection-based adaptive filter configured to;

identify edge pixels and non-edge pixels in the image,

select a filtering technique for at least one non-edge pixel, out of a plurality of filtering techniques for non-edge pixels, based on whether a comparison of the at least one non-edge pixel to a neighboring pixel region indicates that the at least one non-edge pixel is a result of low-light noise, and

generate a transformed image by applying the selected filtering technique to the image.

2. The system of claim 1, wherein the edge-detection-based adaptive filter includes an edge-difference filter configured to divide pixels of the image into the edge pixels and the non-edge pixels.

3. The system of claim 1, wherein the edge-detection-based adaptive filter includes a dilation-based filter configured to expand a region of edge pixels to include non-edge pixels.

4. The system of claim 1, further comprising a Gaussian distribution engine configured to compute a mean and a variance of the Gaussian distribution of the neighboring pixel region.

5. The system of claim 4, wherein the Gaussian distribution engine is further configured to compare the at least one non-edge pixel to the neighboring pixel region by:

determining a difference between a value of the at least one non-edge pixel and the mean of the neighboring pixel region; and

selecting the filtering technique based on a comparison of the difference and the variance of the neighboring pixel region.

6. The system of claim 1, wherein an outcome of the comparison is a numerical value indicating a likelihood of the at least one non-edge pixel being the result of low-light noise, the system further comprising:

a median-filter configured to determine a median value for the neighboring pixel region,

wherein applying the selected filtering technique to the image comprises replacing an original value of the at least one non-edge pixel with the median value when the outcome of the comparison is greater than a first threshold.

7. The system of claim 6, further comprising:

a low-pass filter configured to determine a low-pass filter value for the neighboring pixel region,

wherein applying the selected filtering technique to the image comprises replacing the original value of the at least one non-edge pixel with the low-pass filter value when the outcome of the comparison is less than the first threshold and greater than a second threshold.

8. The system of claim 7, wherein the edge-detection-based adaptive filter is configured to output the low-pass filter value, the median value, or the original value.

9. The system of claim 2, wherein the edge-detection-based adaptive filter includes a dilation-based filter configured to modify an output of the edge-difference filter by distributing results of edge detection to neighboring pixels.

10. A method for removing noise from an image, the method comprising:

identifying edge pixels and non-edge pixels in the image;

selecting a filtering technique for at least one non-edge pixel, out of a plurality of filtering techniques for non-edge pixels, based on whether a comparison of the at least one non-edge pixel to a neighboring pixel region indicates that the at least one non-edge pixel is a result of low-light noise; and

generating a transformed image by applying the selected filtering technique to the image.

11. The method of claim 10, wherein the identifying the edge pixels and non-edge pixels includes expanding a region of edge pixels to include non-edge pixels.

12. The method of claim 10, wherein the selecting the filtering technique for the at least one non-edge pixel based on the comparison of the at least one non-edge pixel to the neighboring pixel region includes:

determining a mean of the neighboring pixel region;

determining a variance of the neighboring pixel region;

determining a difference between an original value of the at least one non-edge pixel and the mean of the neighboring pixel region; and

determining an assigned value of the at least one non-edge pixel based on a comparison of the difference and the variance of the neighboring pixel region.

13. The method of claim 12, wherein an outcome of the comparison is a numerical value indicating a likelihood of the at least one non-edge pixel being the result of low-light noise and wherein the determining the assigned value of the at least one non-edge pixel includes:

when the outcome of the comparison is greater than a first threshold, determining a median value for the neighboring pixel region and assigning the at least one non-edge pixel with the median value;

when the outcome of the comparison is less than the first threshold and greater than a second threshold, determining a low-pass filter value for the neighboring pixel region and assigning the at least one non-edge pixel with the low-pass filter value; and

when the outcome of the comparison is less than the second threshold, assigning the at least one non-edge pixel the original value.

14. The method of claim 12, wherein the determining the mean and the variance includes determining a mean and a variance of a Gaussian distribution of the neighboring pixel region.

15. The method of claim 10, further comprising:

defining a first threshold (N), a second threshold (M), and a third threshold (P), wherein P≦

M≦

N; and

outputting a value for the at least one non-edge pixel based on the comparison of the at least one non-edge pixel to the neighboring pixel region.

16. The method of claim 12, wherein an outcome of the comparison is a numerical value indicating a likelihood of the at least one non-edge pixel being the result of low-light noise and wherein the method further comprises:

determining a median value for the neighboring pixel region and replacing the at least one non-edge pixel with the median value when the outcome of the comparison is greater than the first threshold.

17. The method of claim 10, wherein the neighboring pixel region is a neighboring 3×

3 block of pixels.

21. The method of claim 16, further comprising:

determining a low-pass filter value for the neighboring pixel region and replacing the at least one non-edge pixel with the low-pass filter value when the outcome of the comparison is less than the first threshold and greater than the second threshold.

18. A non-transitory media that includes a program for execution to remove noise from an image and, when executed by a processor, operable to perform operations comprising:

identifying edge pixels and non-edge pixels in the image;

selecting a filtering technique for at least one non-edge pixel, out of a plurality of filtering techniques for non-edge pixels, based on whether a comparison of the at least one non-edge pixel to a neighboring pixel region indicates that the at least one non-edge pixel is a result of low-light noise, and

generating a transformed image by applying the selected filtering technique to the image.

19. The non-transitory media of claim 18, wherein the selecting the filtering technique for the at least one non-edge pixel based on the comparison of the at least one non-edge pixel to the neighboring pixel region includes:

determining a mean of the neighboring pixel region;

determining a variance of the neighboring pixel region;

determining a difference between an original value of the at least one non-edge pixel and the mean of the neighboring pixel region; and

determining an assigned value of the at least one non-edge pixel based on a comparison of the difference and the variance of the neighboring pixel region.

20. The non-transitory media of claim 19, wherein an outcome of the comparison is a numerical value indicating a likelihood of the at least one non-edge pixel being the result of low-light noise and wherein the determining the assigned value of the at least one non-edge pixel includes:

when the outcome of the comparison is greater than a first threshold, determining a median value for the neighboring pixel region and assigning the at least one non-edge pixel with the median value;

when the outcome of the comparison is less than the first threshold and greater than a second threshold, determining a low-pass filter value for the neighboring pixel region and assigning the at least one non-edge pixel with the low-pass filter value; and

when the outcome of the comparison is less than the second threshold, assigning the at least one non-edge pixel the original value.

22. The non-transitory media of claim 18, wherein an outcome of the comparison is a numerical value indicating a likelihood of the at least one non-edge pixel being the result of low-light noise and wherein the operations further comprise:

determining a median value for the neighboring pixel region and replacing the at least one non-edge pixel with the median value when the outcome of the comparison is greater than the first threshold; and

determining a low-pass filter value for the neighboring pixel region and replacing the at least one non-edge pixel with the low-pass filter value when the outcome of the comparison is less than the first threshold and greater than the second threshold.