Method for detecting and correcting defective pixels in a digital image sensor

• US 7,283,164 B2
• Filed: 09/18/2002
• Issued: 10/16/2007
• Est. Priority Date: 09/18/2002
• Status: Active Grant

• Current Assignee: Aptina Imaging Corporation (ON Semiconductor Corporation)
• Original Assignee: Micron Technology, Inc.
• Inventors: Zhang, Xuemei, Ying, Bond Yu-Pong, Stanback, John H., Kakarala, Ramakrishna
• Primary Examiner(s): Vu; Ngoc-Yen
• Assistant Examiner(s): Quiett; Carramah J
• Application Number:
  US10/246,124
• Publication Number:
  US 20040051798A1
• Time in Patent Office: 1,854 Days
• Field of Search: 348/243, 348/246, 348/247, 348/272, 348/250, 382/162, 382/250, 382/167, 382/172
• US Class Current: 348/246
• CPC Class Codes: H04N 2209/045 using mosaic colour filter
• CPC Class Codes: H04N 25/68 applied to defects

Abstract

A bad pixel correction (BPC) algorithm that can be implemented on the image sensor chip is provided for detecting and correcting defective pixels in a digital color image sensor. Gradients of neighboring pixels in at least one other color plane than the color plane of a current pixel and a range of sensor values from neighboring pixels in the same color plane as the current pixel are determined. If the sensor value of the current pixel is outside of the range by a threshold amount that is calculated using one or more of the gradients, the current pixel is determined to be a defective pixel, and replaced using the sensor values of the neighboring pixels in the same color plane.

Claims

1. A digital image sensor for detecting and correcting sensor values from defective pixels in an image, comprising:

an array of pixels arranged in rows and columns, each of said pixels in said array of pixels producing a respective sensor value for an image;

a buffer for storing at least part of said sensor values produced by at least one of said rows of said pixels for said image, one of said sensor values being a current sensor value in a first color plane produced by a current one of said pixels; and

a processor connected;

(1) to receive said sensor values stored in said buffer, (2) to compute one or more gradients from said sensor values in at least a second color plane and (3) to calculate a range of values corresponding to said current pixel based on said sensor values produced by neighboring ones of said pixels in said first color plane and a threshold amount using said one or more gradients and said range of values, the range of values including at least a maximum range value and a minimum range value,

wherein the processor determines whether said current pixel is defective responsive to the current pixel value being one of;

(i) greater than said maximum range value of the range of values by more than said threshold amount or (ii) less than said minimum range value of the range of values by more than said threshold amount.

2. The sensor of claim 1, wherein said processor replaces said current sensor value with a replacement value calculated using said sensor values produced by said neighboring pixels.

3. The sensor of claim 2, wherein said range of values includes a maximum sensor value, as the maximum range value, from said sensor values produced by said neighboring pixels and a minimum sensor value, as a minimum range value, from said sensor values produced by said neighboring pixels, said current pixel being determined to be defective responsive to an absolute value of a difference between said current sensor value and either said maximum range value or said minimum range value exceeding said threshold amount.

4. The sensor of claim 3, wherein said replacement value is a median value of said sensor values produced by said neighboring pixels.

5. The sensor of claim 3, wherein said maximum range value is increased by a maximum of one or more same color gradients computed from said neighboring pixels and said minimum range value is decreased by a minimum of said one or more same color gradients when said current sensor value is consistent with said one or more same color gradients.

6. The sensor of claim 5, wherein said one or more same color gradients includes one or more current row same color gradients computed from said neighboring pixels in a current one of said rows including said current pixel.

7. The sensor of claim 6, wherein said one or more same color gradients further includes one or more vertical same color gradients computed from said neighboring pixels in said current row and one or more additional ones of said rows above or below said current row, said maximum range value being increased by a maximum of said current row same color gradients and said minimum range value being decreased by a minimum of said current row same color gradients.

8. The sensor of claim 3, wherein said maximum range value is increased by a maximum of said one or more gradients computed from additional neighboring ones of said pixels in said second color plane within a current one of said rows including said current pixel and said minimum range value is decreased by a minimum of said one or more gradients computed from said additional neighboring pixels when said current sensor value is consistent with said one or more gradients.

9. The sensor of claim 3, wherein said processor further calculates a minimum average gradient using said one or more gradients, said threshold amount being calculated using said minimum average gradient and one or more threshold parameters.

10. The sensor of claim 9, wherein said minimum average gradient is a minimum of an absolute value of an average of said one or more gradients in said first color plane and said second plane on a first side of said current pixel and an absolute value of an average of said one or more gradients in said first color plane and said second color plane on a second side of said current pixel.

11. The sensor of claim 9, wherein said one or more threshold parameters include a scale factor multiplied by said minimum average gradient and a pedestal added to a product of said minimum average gradient and said scale factor.

12. The sensor of claim 9, wherein said processor further calculates a minimum vertical gradient from one or more vertical gradients computed from respective neighboring ones of said pixels to said current pixel in said first color plane in different ones of said rows than said current pixel.

13. The sensor of claim 12, wherein said processor determines that said current pixel is defective when said difference exceeds said threshold amount and said minimum vertical gradient exceeds said threshold amount.

14. The sensor of claim 3, wherein said processor further determines that said current pixel is defective when said current sensor value exceeds said maximum range value by more than a threshold parameter.

15. The sensor of claim 2, wherein said range of values includes a maximum estimate, a minimum estimate and a mean estimate calculated using said one or more gradients, said threshold amount being calculated using said mean estimate and one or more threshold parameters, said current pixel being defective when said current sensor value is greater than said maximum estimate or less than said minimum estimate by more than said threshold amount.

16. The sensor of claim 15, wherein said replacement value is said mean estimate.

17. The sensor of claim 15, wherein said range of values includes eight estimates, said maximum estimate being a maximum of said eight estimates, said minimum estimate being a minimum of said eight estimates and said mean estimate being a mean of said eight estimates, each of said eight estimates being calculated by adding said sensor value from a surrounding one of said pixels in said first color plane to said one or more gradients in at least said second color plane in a direction of said added sensor value.

18. A method for detecting defective pixels, comprising:

storing sensor values produced by at least part of one row of pixels in a sensor for an image, one of said sensor values being a current sensor value in a first color plane produced by a current one of said pixels;

computing one or more gradients from said sensor values in at least a second color plane;

calculating a range of values corresponding to said current pixel based on said sensor values produced by neighboring ones of said pixels in said first color plane;

calculating said threshold amount using said one or more gradients; and

determining whether said current pixel is defective responsive to the current pixel value being one of;

(i) greater than a maximum range value of the range of values by more than said threshold or (ii) less than a minimum range value of the range of values by more than said threshold amount.

19. The method of claim 18, further comprising:

replacing said current sensor value with a replacement value calculated using said sensor values produced by said neighboring pixels.

20. The method of claim 19, wherein said range of values includes a maximum sensor value, as the maximum range value, and a minimum sensor value, as the minimum range value, from said sensor values produced by said neighboring pixels, said step of determining further comprising:

determining that said current pixel is defective responsive to an absolute value of a difference between said current sensor value and either said maximum range value or said minimum range value exceeding said threshold amount.

21. The method of claim 20, wherein said step of replacing further comprises:

calculating a median value of said sensor values produced by said neighboring pixels; and

using said median value as said replacement value.

22. The method of claim 20, wherein said step of calculating said range of values further comprises:

increasing said maximum range value by a maximum of one or more same color gradients computed from said neighboring pixels when said current sensor value is consistent with said one or more same color gradients; and

decreasing said minimum range value by a minimum of said one or more same color gradients when said current sensor value is consistent with said one or more same color gradients.

23. The method of claim 22, wherein said step of calculating said range of values further comprises:

computing said one or more same color gradients from said neighboring pixels in a current one of said rows including said current pixel.

24. The method of claim 23, wherein said step of calculating said range of values further comprises:

computing said one or more same color gradients from said neighboring pixels in said current row and one or more additional ones of said rows above or below said current row.

25. The method of claim 20, wherein said step of calculating said range of values further comprises:

increasing said maximum range value by a maximum of said one or more gradients computed from additional neighboring ones of said pixels in said second color plane within a current one of said rows including said current pixel when said current sensor value is consistent with said one or more gradients; and

decreasing said minimum range value by a minimum of said one or more gradients computed from said additional neighboring pixels when said current sensor value is consistent with said one or more gradients.

26. The method of claim 20, wherein said step of calculating said threshold amount further comprises:

calculating a minimum average gradient using said one or more gradients; and

calculating said threshold amount using said minimum average gradient and one or more threshold parameters.

27. The method of claim 26, wherein said step of calculating said minimum average gradient further comprises:

determining a minimum of a absolute value of an average of said one or more gradients in said first color plane and said second plane on a first side of said current pixel and an absolute value of a average of said one or more gradients in said first color plane and said second color plane on a second side of said current pixel.

28. The method of claim 26, wherein the one or more threshold parameters include a scale factor and a pedestal, and wherein said step of calculating said threshold amount further comprises:

multiplying said scale factor by said minimum average gradient; and

adding said pedestal to a product of said minimum average gradient and said scale factor.

29. The method of claim 26, wherein said step of determining that said current pixel is defective further comprises:

calculating a minimum vertical gradient from one or more vertical gradients computed from respective neighboring ones of said pixels to said current pixel in said first color plane in different ones of said rows than said current pixel; and

determining that said current pixel is defective when said difference exceeds said threshold amount and said minimum vertical gradient exceeds said threshold amount.

30. The method of claim 20, wherein said step of determining that said current pixel is defective further comprises:

determining that said current pixel is defective when said current sensor value exceeds said maximum range value by more than a threshold parameter.

31. The method of claim 19, wherein said range of values includes a maximum estimate, a minimum estimate and a mean estimate calculated using said one or more gradients, said threshold amount being calculated using said mean estimate and one or more threshold parameters, said step of determining further comprising:

determining that said current pixel is defective when said current sensor value is greater than said maximum estimate or less than said minimum estimate by more than said threshold amount.

32. The method of claim 31, wherein said step of replacing further comprises:

replacing said current sensor value with said mean estimate.

33. The method of claim 31, wherein said range of values includes eight estimates, said maximum estimate being a maximum of said eight estimates, said minimum estimate being a minimum of said eight estimates and said mean estimate being a mean of said eight estimates, said step of calculating said range of values further comprising:

calculating each of said eight estimates by adding said sensor value from a surrounding one of said pixels in said first color plane to said one or more gradients in said second color plane in a direction of said added sensor value.