Noise filter for bayer pattern image data

• US 20030048368A1
• Filed: 08/28/2002
• Published: 03/13/2003
• Est. Priority Date: 08/31/2001
• Status: Active Grant

• Current Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Original Assignee: STMicroelectronics SRL (STMicroelectronics NV)
• Inventors: Bosco, Angelo, Mancuso, Massimo
• Granted Patent: US 7,369,165 B2
• Time in Patent Office: Days
• US Class Current: 348/272
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20012 Locally adaptive
• CPC Class Codes: G06T 2207/20192 Edge enhancement; Edge preservation
• CPC Class Codes: G06T 3/4015 Image demosaicing, e.g. colour filter arrays [CFA] or Bayer patterns
• CPC Class Codes: G06T 5/002 Denoising; Smoothing noise processing or correction adapted to be used in an image pickup device containing and electronic image sensor H04N23/81, H04N25/60 - H04N25/67
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/70 Denoising; Smoothing
• CPC Class Codes: H04N 23/81 for suppressing or minimising disturbance in the image signal generation
• CPC Class Codes: H04N 23/843 Demosaicing, e.g. interpolating colour pixel values
• CPC Class Codes: H04N 25/134 based on three different wavelength filter elements

Abstract

A method of filtering an image filter is disclosed. The filter is provided for a digital camera including image sensors sensitive to light, a color filter placed over sensitive elements of the sensors and patterned according to a Bayer mosaic pattern layout and an interpolation algorithm joining together the digital information provided by differently colored adjacent pixels in said Bayer pattern. The filter is adaptive and includes a noise level computation block for operating directly on a said Bayer pattern data set of for each color channel thus removing noise while simultaneously preserving picture detail.

Claims

1. An image filter for filtering a Bayer pattern, comprising:

an interpolation algorithm that joins together the digital information provided by differently colored adjacent pixels in a plurality of color channel Bayer patterns; and

the filter being adaptive and operable directly on the Bayer patterns for each color channel to remove noise while simultaneously preserving picture detail.

2. The image filter according to claim 1 wherein at least two filter masks are used, one for a green channel and the other for a red and blue channels.

3. The image filter according to claim 2 wherein the filter masks for the red and blue channels are identical while the filter mask for the green channel is different.

4. The image filter according to claim 1 further including:

a determination function that examines a current pixel in a processing window, determines its color, and selects an appropriate filtering mask for that color;

a difference calculator that determines an absolute value difference between said current pixel and the values of a plurality of its neighboring similarly colored pixels in said processing window;

a maximum and minimum determination function that computes a maximum and minimum one of the determined distance values; and

a noise level computation function that estimates a noise level associated with said processing window.

5. The image filter according to claim 4 further including a gain factor function comprising an autoexposure control system that generates an appropriate set of exposure and gain settings fed to said noise level computation function.

6. The image filter according to claim 4 wherein said noise level computation function uses a recursive implementation and estimates the noise level according to the following calculations:

NLR(t−

1)=Kn(t−

1)*Dmax(t−

1)+[1−

Kn(t−

1)]*NLR(t−

2);

NLG(t−

1)=Kn(t−

1)*Dmax(t−

1)+[1−

Kn(t−

1)]*NLG(t−

2);

NLB(t−

1)=Kn(t−

1)*Dmax(t−

1)+[1−

Kn(t−

1)]*NLB(t−

2)

where NLx is a noise level of the particular noise channel, x;

Kn is a parameter that determines a strength of filtering to be performed by the filter, and Dmax is the maximum distance value.

7. A digital camera, comprising:

a sensor for sensing an image and producing a first signal;

a Bayer pattern producer coupled to said sensor and structured to produce a Bayer pattern from said first signal;

a splitter structured to split said Bayer pattern into separate color channels; and

an image filter for processing each of the separate color channels of the Bayer pattern in a manner that removes noise while simultaneously preserving picture detail.

8. A method for filtering noise in a digital image, comprising the steps of:

interpolating the digital image to join together digital information provided by differently colored adjacent pixels and produce a plurality of color channel Bayer patterns;

scanning sequentially row by row through a selected Bayer pattern;

providing a processing window for each color channel of said Bayer pattern data set, the processing windows for the red and blue channels being identical, while the processing window for the green channel being different, and each processing window having a target pixel and a plurality of neighboring pixels;

selecting a filtering mask and a corresponding processing window in dependence on color of a current target pixel; and

filtering the target pixel by computing a noise level of the color channel and providing a filtered pattern format depending on said noise level.

9. The method according to claim 8, wherein the filtered value of the current target pixel is computed using the pixel values that have already been filtered during said scanning step.

10. An image filtering method for processing a color channel of Bayer pattern data, comprising the steps of:

identifying a target pixel in the Bayer pattern data;

identifying, within the Bayer pattern data, a plurality of neighboring pixels to the target pixel having a common color;

comparing the common color target and neighboring pixels; and

determining a filtered value for the target pixel based on the common color target and neighboring pixels comparison by;

resisting change in the target pixel value when the comparison indicates dramatic difference between the target and neighboring pixels; and

encouraging change in the target pixel value when the comparison indicates minimal difference between the target and neighboring pixels.

11. The method as in claim 10 wherein the comparing step implements a fuzzy logic processing operation.

12. The method as in claim 10 wherein the step of resisting comprises the step of reducing a weighting factor applied to each of the neighboring pixel values if the comparison indicates dramatic difference between the target and neighboring pixels.

13. The method as in claim 10 wherein the step of encouraging comprises the step of increasing a weighting factor applied to each of the neighboring pixel values if the comparison indicates minimal difference between the target and neighboring pixels.

14. The method as in claim 10 wherein the step of determining comprises the step of calculating the filtered value of the target pixel as a sum of:

a weighting factor times the neighboring pixel value; and

one minus the weighting factor times the target pixel value;

wherein the sum is calculated across the plurality of neighboring pixels.

15. An image filter for processing a color channel of Bayer pattern data, comprising:

a working window function that identifies a target pixel in the Bayer pattern data along with a plurality of neighboring pixels to the target pixel having a common color;

a comparator function that compares the common color target and neighboring pixels; and

a filtering function that determines a filtered value for the target pixel based on the common color target and neighboring pixels comparison by;

resisting change in the target pixel value when the comparison indicates dramatic difference between the target and neighboring pixels; and

encouraging change in the target pixel value when the comparison indicates minimal difference between the target and neighboring pixels.

16. The filter as in claim 15 wherein the comparator function implements a fuzzy logic processing operation.

17. The filter as in claim 15 wherein the filtering function resists change by reducing a weighting factor applied to each of the neighboring pixel values if the comparison indicates dramatic difference between the target and neighboring pixels.

18. The filter as in claim 15 wherein the filtering function encourages change by increasing a weighting factor applied to each of the neighboring pixel values if the comparison indicates minimal difference between the target and neighboring pixels.

19. The filter as in claim 15 wherein the filtering function calculates the filtered value of the target pixel as a sum of:

a weighting factor times the neighboring pixel value; and

one minus the weighting factor times the target pixel value;

wherein the sum is calculated across the plurality of neighboring pixels.

20. An image filtering method for processing a color channel of Bayer pattern data, comprising the steps of:

identifying a target pixel in the Bayer pattern data;

identifying, within the Bayer pattern data, a plurality of neighboring pixels to the target pixel having a common color;

comparing the common color target and neighboring pixels; and

determining a filtered value for the target pixel based on the common color target and neighboring pixels comparison that;

preserves image detail where the comparison indicates dramatic difference between the target and neighboring pixels; and

corrects noisy pixels where the comparison indicates minimal difference between the target and neighboring pixels.

21. The method as in claim 20 wherein the comparing step implements a fuzzy logic processing operation.

22. The method as in claim 20 wherein the step of preserving comprises the step of resisting change in the target pixel value by reducing a weighting factor applied to each of the neighboring pixel values if the comparison indicates dramatic difference between the target and neighboring pixels.

23. The method as in claim 20 wherein the step of correcting comprises the step of encouraging change in the target pixel value by increasing a weighting factor applied to each of the neighboring pixel values if the comparison indicates minimal difference between the target and neighboring pixels.

24. The method as in claim 20 wherein the step of determining comprises the step of calculating the filtered value of the target pixel as a sum of:

a weighting factor times the neighboring pixel value; and

one minus the weighting factor times the target pixel value;

wherein the sum is calculated across the plurality of neighboring pixels.

25. An image filter for processing a color channel of Bayer pattern data, comprising:

a working window function that identifies a target pixel in the Bayer pattern data along with a plurality of neighboring pixels to the target pixel having a common color;

a comparator function that compares the common color target and neighboring pixels; and

a filtering function that determines a filtered value for the target pixel based on the common color target and neighboring pixels comparison that;

preserves image detail where the comparison indicates dramatic difference between the target and neighboring pixels; and

corrects noisy pixels where the comparison indicates minimal difference between the target and neighboring pixels.

26. The filter as in claim 25 wherein the comparator function implements a fuzzy logic processing operation.

27. The filter as in claim 25 wherein the filtering function preserves image detail by resisting change in the target pixel value by reducing a weighting factor applied to each of the neighboring pixel values if the comparison indicates dramatic difference between the target and neighboring pixels.

28. The filter as in claim 25 wherein the filtering function corrects noisy pixels by encouraging change by increasing a weighting factor applied to each of the neighboring pixel values if the comparison indicates minimal difference between the target and neighboring pixels.

29. The filter as in claim 25 wherein the filtering function calculates the filtered value of the target pixel as a sum of:

a weighting factor times the neighboring pixel value; and

one minus the weighting factor times the target pixel value;

wherein the sum is calculated across the plurality of neighboring pixels.