Covariance based color characteristics of images

• US 8,971,668 B2
• Filed: 02/27/2013
• Issued: 03/03/2015
• Est. Priority Date: 02/27/2013
• Status: Active Grant

• Current Assignee: Adobe Inc.
• Original Assignee: Adobe Systems Incorporated (Adobe Inc.)
• Inventors: Brandt, Jonathan W., Becherer, Nico A., Duwenhorst, Sven
• Primary Examiner(s): Bayat, Ali
• Application Number:
  US13/778,938
• Publication Number:
  US 20140244669A1
• Time in Patent Office: 734 Days
• Field of Search: 382/103, 382/159, 382/115, 382/145, 382/162, 382/165, 382/190, 382/236, 382/305, 707/754, 706/52
• US Class Current: 382/305
• CPC Class Codes: G06F 16/5838 using colour
• CPC Class Codes: G06V 10/56 relating to colour

Abstract

Each of multiple images is analyzed to determine how the colors of the pixels of the image are distributed throughout the color space of the image. Different covariance based characteristics of the image are determined that identify a direction, as well as magnitude in each direction, of the distribution of colors of the image pixels. These different covariance based characteristics that are determined for an image can be saved as associated with the image, allowing the characteristics to be accessed and used as a basis for searching the images to identify particular types of images. These different covariance based characteristics can also be used to order the images identified by a search.

Claims

1. A method comprising:

receiving a request to search multiple images, the request being associated with particular search criteria;

obtaining covariance based characteristics of each of the multiple images, the covariance based characteristics of an image identifying one or both of a direction and a magnitude, in a color space of the image, of a distribution of color values of pixels of the image;

identifying, based on both the covariance based characteristics of the multiple images and the particular search criteria, one or more of the multiple images that satisfy the search criteria; and

providing, in response to the request, the one or more of the multiple images.

2. A method as recited in claim 1, further comprising ordering the one or more of the multiple images in accordance with the covariance based characteristics.

3. A method as recited in claim 1, the obtaining the covariance based characteristics comprising retrieving the covariance based characteristics stored as associated with the image.

4. A method as recited in claim 1, the obtaining the covariance based characteristics comprises generating, for at least one of the multiple images, the covariance based characteristics in response to the request.

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

obtaining, for each of the multiple images, mean color values for the image that identify the mean color values for pixels of the image in each element of the color space; and

the identifying comprising identifying, based on the covariance based characteristics of the multiple images as well as the mean color values for the images and the particular search criteria, one or more of the multiple images that satisfy the search criteria.

6. A method as recited in claim 1, the obtaining comprising obtaining, as covariance based characteristics identifying the magnitude of the distribution of color values of pixels of the image, one or more eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space.

7. A method as recited in claim 1, the obtaining further comprising obtaining, as covariance based characteristics of the image, a dominant color angle of the image, the dominant color angle comprising an angle between a grayscale vector in the color space and a dominant eigenvector of the image, the dominant eigenvector of the image comprising an eigenvector of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space, the dominant eigenvector corresponding to a largest of multiple eigenvalues of the covariance matrix.

8. A method as recited in claim 7, the receiving the request comprising receiving a request to search for grayscale images, the request being associated with search criteria indicating:

one of the multiple eigenvalues being less than a first threshold amount, the one eigenvalue comprising the one of the multiple eigenvalues having a second largest magnitude; and

the dominant color angle being less than a second threshold amount.

9. A method as recited in claim 7, the receiving the request comprising receiving a request to search for sepia images, the request being associated with search criteria indicating:

one of the multiple eigenvalues being less than a first threshold amount, the one eigenvalue comprising the one of the multiple eigenvalues having a second largest magnitude, the first threshold amount being greater than a second threshold amount included as search criteria for the one eigenvalue associated with requests to search for grayscale images; and

the dominant color angle being less than a third threshold amount, the third threshold amount being less than a fourth threshold amount included as search criteria for the dominant color angle associated with requests to search for grayscale images.

10. A method as recited in claim 1, the receiving the request comprising receiving a request to search for low contrast images, the request being associated with search criteria indicating one of multiple eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space being less than a first threshold amount, the one eigenvalue being the one of the multiple eigenvalues having the largest magnitude.

11. A method as recited in claim 1, the receiving the request comprising receiving a request to search for high contrast images, the request being associated with search criteria indicating one of multiple eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space being greater than a first threshold amount, the one eigenvalue being the one of the multiple eigenvalues having the largest magnitude, the first threshold amount being greater than a second threshold amount included as search criteria for the one eigenvalue associated with requests to search for low contrast images.

12. A method as recited in claim 1, the receiving the request comprising receiving a request to search for low color images, the request being associated with search criteria indicating one of multiple eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space being less than a first threshold amount, the one eigenvalue being the one of the multiple eigenvalues having a second largest magnitude.

13. A method as recited in claim 1, the receiving the request comprising receiving a request to search for medium color images, the request being associated with search criteria indicating:

a first eigenvalue of multiple eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space being at least a first threshold amount, the first eigenvalue being the one of the multiple eigenvalues having a second largest magnitude; and

a second eigenvalue of the multiple eigenvalues being less than a second threshold amount, the second eigenvalue being the one of the multiple eigenvalues having a smallest magnitude.

14. A method as recited in claim 1, the receiving the request comprising receiving a request to search for high color images, the request being associated with search criteria indicating:

a first eigenvalue of multiple eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space being at least a first threshold amount, the first eigenvalue being the one of the multiple eigenvalues having a second largest magnitude; and

a second eigenvalue of the multiple eigenvalues being at least a second threshold amount, the second eigenvalue being the one of the multiple eigenvalues having a smallest magnitude.

15. A method comprising:

determining, for each of multiple images, covariance based characteristics of the image upon which image searches by an image searching system can be based, the covariance based characteristics of the image identifying one or both of a direction and a magnitude, in a color space of the image, of a distribution of color values of pixels of the image; and

storing the covariance based characteristics of the multiple images so that the covariance based characteristics of each of the multiple images is accessible to the image searching system.

16. A method as recited in claim 15, further comprising:

determining, for each of the multiple images, mean color values for the image that identify the mean color values for pixels of the image in each element of the color space; and

the storing comprising storing both the covariance based characteristics and the mean color values so that both the covariance based characteristics and the mean color values of each of the multiple images are accessible to the image searching system.

17. A method as recited in claim 16, the storing comprising storing both the covariance based characteristics and the mean color values for an image as associated with the image.

18. A method as recited in claim 15, the determining comprising determining, as the covariance based characteristics identifying one or both of the direction and the magnitude of the distribution of color values of pixels of the image, one or more eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space.

19. A method as recited in claim 18, the determining further comprising determining, as covariance based characteristics of the image, a dominant color angle of the image, the dominant color angle comprising an angle between a grayscale vector in the color space and a dominant eigenvector of the image, the dominant eigenvector of the image comprising an eigenvector of the covariance matrix, the dominant eigenvector corresponding to a largest of multiple eigenvalues of the covariance matrix.

20. A device comprising:

a memory to store covariance based characteristics of each of multiple images, the covariance based characteristics of an image identifying a magnitude, in each of one or more directions in a color space of the image, of a distribution of color values of pixels of the image; and

a processing system to implement a covariance based image searching system configured to;

obtain, in response to a search request associated with search criteria, the covariance based characteristics of each of the multiple images;

use the covariance based characteristics of the multiple images to identify one or more of the multiple images that satisfy the search criteria; and

provide, in response to the search request, the identified one or more images.

21. A device as recited in claim 20, the covariance based image searching system being further configured to order the identified one or more images in accordance with the covariance based characteristics of the identified one or more images.

22. A device as recited in claim 20, wherein to obtain the covariance based characteristics is to retrieve the covariance based characteristics stored as associated with the image.

23. A device as recited in claim 20, the covariance based image searching system being further configured to obtain, as covariance based characteristics of an image, one or more eigenvalues of a covariance matrix that indicates how the colors of pixels in the image are spread out in the color space.

24. A device as recited in claim 23, the covariance based image searching system being further configured to obtain, as covariance based characteristics of an image, a dominant color angle of the image, the dominant color angle comprising an angle between a grayscale vector in the color space and a dominant eigenvector of the image, the dominant eigenvector of the image comprising an eigenvector of the covariance matrix, the dominant eigenvector corresponding to a largest of multiple eigenvalues of the covariance matrix.