Dynamic range correction based on image content

• US 8,447,132 B1
• Filed: 12/06/2010
• Issued: 05/21/2013
• Est. Priority Date: 12/09/2009
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: CSR Technology Incorporated (Qualcomm, Inc.)
• Inventors: Galil, Erez
• Primary Examiner(s): PATEL, KANJIBHAI B
• Application Number:
  US12/961,400
• Time in Patent Office: 897 Days
• Field of Search: 382/118, 382/243, 382/254, 382/264, 382/266, 382/274, 382/305, 382/312, 382/275, 348/207.99, 348/241, 348/345, 358/1.9, 358/3.27, 358/518, 358/520, 345/617
• US Class Current: 382/274
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation
• CPC Class Codes: H04N 1/4074 using histograms

Abstract

Disclosed are method and a corresponding apparatus, where the method according to one embodiment includes making a determination that a first portion of digital image data represents a physical object of a predetermined type, determining an amount of a parameter, such as a gain, to apply to the first portion of the digital image data, based on the determination that the first portion of the digital image data represents a physical object of the predetermined type, and applying the determined amount of the parameter to the first portion of the digital image data. The method may be part of a dynamic range correction operation.

Claims

1. A method comprising:

determining whether a first portion of digital image data represents a physical object of a predetermined type;

determining a correction to apply to the first portion of the digital image data, based on a determination that the first portion of the digital image data represents a physical object of the predetermined type, wherein the determined correction is matched to the predetermined type;

applying the determined correction to the first portion of the digital image data to enhance a visual characteristic of the first portion of the digital image data, by applying a first amount of the correction to the first portion of the digital image data; and

applying a second amount of the correction to a second portion of the digital image data, wherein the first amount differs from the second amount, and wherein the first amount corresponds to a physical object of the predetermined type.

2. A method as recited in claim 1, wherein the correction comprises a dynamic range correction.

3. A method as recited in claim 1, wherein determining the correction comprises determining an amount of a parameter to apply to the first portion of the digital image data.

4. A method as recited in claim 3, wherein the parameter is a gain.

5. A method as recited in claim 1, wherein applying the determined correction to the first portion of the digital image data comprises applying a first amount of a parameter to the first portion to enhance the visual characteristic, wherein the first amount of the parameter is different from a second amount of the parameter applied to a second portion of the digital image data, resulting in a first portion of a digital image that corresponds to the first portion of digital image data being brighter than if the second amount of the parameter were applied to both the first portion and the second portion of the digital image data.

6. A method as recited in claim 1, wherein the predetermined type of physical object represents facial attributes.

7. A method as recited in claim 1, further comprising:

determining that the second portion of the digital image data does not represent a physical object of the predetermined type; and

determining a different amount of gain to apply to the second portion than an amount of gain determined for the first portion.

8. A method as recited in claim 7, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

determining the amount of gain to apply to the first portion of the digital image data so as to maintain local contrast of the first portion of the digital image data.

9. A method as recited in claim 7, wherein the first portion of the digital image data includes a plurality of pixels, and wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

determining a first intermediate value based on a measure of image strength of a first pixel within the first portion of the digital image data;

determining a second intermediate value based on a measure of image strength of a plurality of pixels surrounding the first pixel in the digital image data; and

using the first intermediate value and the second intermediate value to determine the amount of gain to apply to the first portion of the digital image data.

10. A method as recited in claim 9, wherein determining the first intermediate value and determining the second intermediate value each separately comprise applying a low pass filter function.

11. A method as recited in claim 7, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

selecting a value from a gain data structure according to whether the first portion of the digital image data represents a physical object of the predetermined type.

12. A method as recited in claim 7, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

determining a first gain based on the determination that the first portion of the digital image data represents a physical object of the predetermined type; and

wherein applying the determined amount of gain comprises applying the first gain to a pixel value from the first portion of the digital image data and a second gain, wherein the second gain is independent of the determination that the first portion of the digital image data represents a physical object of the predetermined type.

13. A method as recited in claim 7, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

modifying a gain map based on the determination that the first portion of the digital image data represents a physical object of the predetermined type; and

wherein applying the determined amount of gain to apply to the first portion of the digital image data comprises obtaining the amount of gain to apply to the first portion of the digital image data from the gain map.

14. A method as recited in claim 7, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

selecting one of a plurality of data structures of gain values, from which to obtain a gain value, according to whether the first portion of the digital image data represents a physical object of the predetermined type.

15. A method comprising:

making a determination, in a digital image processing device, that a first group of adjacent pixels in a digital image represents a face;

determining, in the digital image processing device, a type or amount of dynamic range correction to apply to at least one pixel in the first group of adjacent pixels based on the determination that the first group of adjacent pixels represents a face, wherein said determining includes determining a different amount of gain for the first group of pixels than that determined for a second group of pixels which do not represent a face, wherein said determining further includes determining the amount of gain for the first group of adjacent pixels so as to maintain local contrast of the first group of adjacent pixels, and wherein said determining further includes

determining a first intermediate value based on a measure of image strength of a first pixel of the first group of adjacent pixels;

determining a second intermediate value based on a measure of image strength of a plurality of pixels surrounding the first pixel; and

using the first intermediate value and the second intermediate value to determine the type or amount of gain;

applying the determined type or amount of dynamic range correction to the at least one pixel in the first group of adjacent pixels;

determining that a second group of pixels in the digital image does not represent a face; and

determining a different type or amount of dynamic range correction to apply to the second group of pixels than that determined for the first group.

16. A method as recited in claim 15, wherein determining the first intermediate value and determining the second intermediate value each separately comprise applying a low pass filter function.

17. A method as recited in claim 15, wherein determining a type or amount of dynamic range correction to apply to at least one pixel in the first group of adjacent pixels comprises:

selecting a value from a gain data structure according to whether the at least one pixel is part of a portion of image data that represents a physical object of the predetermined type.

18. A method as recited in claim 15, wherein determining a type or amount of dynamic range correction to apply to at least one pixel in the first group of adjacent pixels comprises:

determining a first gain based on the determination that the at least one pixel represents a physical object of the predetermined type; and

wherein applying the determined type or amount of dynamic range correction comprises applying the first gain to a pixel value from the first portion of the digital image data and a second gain, wherein the second gain is independent of the determination that the first portion of the digital image data represents a physical object of the predetermined type.

19. A method as recited in claim 15, wherein determining a type or amount of dynamic range correction to apply to at least one pixel in the first group of adjacent pixels comprises:

modifying a gain map based on the determination that the first portion of the digital image data represents a physical object of the predetermined type; and

wherein applying the determined type or amount of dynamic range correction comprises obtaining the amount of gain from the gain map.

20. A method as recited in claim 15, wherein determining a type or amount of dynamic range correction to apply to at least one pixel in the first group of adjacent pixels comprises:

selecting one of a plurality of data structures of gain values, from which to obtain a gain value, according to whether the at least one pixel is part of a portion of image data that represents a physical object of the predetermined type.

21. An image processing unit comprising:

an object detection unit to receive an intensity value for each of a plurality of pixels in a set of digital image data and to determine that a first portion of the set of digital image data represents a physical object of a predetermined type; and

a correction unit to determine a type or amount of correction to apply to the first portion of the set of digital image data, based on an output of the object detection unit, and to apply the determined type or amount of correction to the first portion of the set of digital image data, and to apply a different type or amount of correction to a second portion of the set of digital image data which does not represent a physical object of the predetermined type.

22. An image processing unit as recited in claim 21, wherein the correction is a dynamic range correction.

23. An image processing unit as recited in claim 22, wherein the object detection unit is configured to detect faces.

24. An image processing unit as recited in claim 21, wherein the correction unit further is to:

determine that a second portion of the set of digital image data does not represent a physical object of the predetermined type; and

determine a different type or amount of correction to apply to the second portion than that determined for the first portion.

25. An image processing unit as recited in claim 21, wherein the correction unit comprises:

a filter unit which receives the intensity values of the plurality of pixels and which includes a plurality of cascaded filter stages, wherein at least one filter stage of the plurality of cascaded filter stages operates based on a larger pixel neighborhood than a preceding filter stage of the plurality of cascaded filter stages, and wherein the image processing unit determines a number of stages of the filter unit that are active based on the set of digital image data;

wherein the correction unit corrects intensity values of the plurality of pixels based on outputs of the filter unit.

26. An image processing unit as recited in claim 21, wherein the correction unit determines an amount of gain to apply to the first portion of the digital image data by determining the amount of gain so as to maintain local contrast of the first portion of the digital image data.

27. An image processing unit as recited in claim 26, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

selecting a value from a gain data structure according to whether the first portion of the digital image data represents a physical object of the predetermined type.

28. An image processing unit as recited in claim 26, wherein the first portion of the digital image data includes a plurality of pixels, and wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

determining a first intermediate value based on a measure of image strength of a first pixel within the first portion of the digital image data;

determining a second intermediate value based on a measure of image strength of a plurality of pixels surrounding the first pixel in the digital image data; and

using the first intermediate value and the second intermediate value to determine the amount of gain.

29. An image processing unit as recited in claim 28, wherein determining the first intermediate value and determining the second intermediate value each separately comprise applying a low pass filter function.

30. An image processing unit as recited in claim 26, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

modifying a gain data structure based on the determination that the first portion of the digital image data represents a physical object of the predetermined type; and

wherein applying the determined amount of gain comprises obtaining the amount of gain from the gain data structure.

31. An image processing unit as recited in claim 26, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

determining a first gain based on the determination that the first portion of the digital image data represents a physical object of the predetermined type; and

wherein applying the determined amount of gain comprises applying the first gain to a pixel value from the first portion of the digital image data and a second gain, wherein the second gain is independent of the determination that the first portion of the digital image data represents a physical object of the predetermined type.

32. An image processing unit as recited in claim 26, wherein determining the amount of gain to apply to the first portion of the digital image data comprises:

selecting one of a plurality of data structures of gain values, from which to obtain a gain value, according to whether the first portion of the digital image data represents a physical object of the predetermined type.

33. An apparatus comprising:

means for making a determination that a first group of adjacent pixels in a digital image represents a physical object of a predetermined type;

means for determining a type or amount of dynamic range correction to apply to at least one pixel in the first group of adjacent pixels based on the determination that the first group of adjacent pixels represents a physical object of the predetermined type;

means for applying the determined type or amount of dynamic range correction to the at least one pixel in the first group of adjacent pixels, by applying a first amount of the correction to the first portion of the digital image data; and

means for applying a second amount of the correction to a second portion of the digital image data, wherein the first amount differs from the second amount, and wherein the first amount corresponds to a physical object of the predetermined type.