Method and system for generating weights for use in white balancing an image

• US 9,826,208 B2
• Filed: 06/26/2013
• Issued: 11/21/2017
• Est. Priority Date: 06/26/2013
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: NVIDIA Corporation
• Inventors: Cabral, Brian, Motta, Ricardo, Harwell, Mitchell
• Primary Examiner(s): Peterson, Christopher K
• Application Number:
  US13/927,405
• Publication Number:
  US 20150002692A1
• Time in Patent Office: 1,609 Days
• CPC Class Codes: G06F 18/41 Interactive pattern learning with a human teacher
• CPC Class Codes: H04N 23/88 for colour balance, e.g. white-balance circuits or colour temperature control

Abstract

Embodiments of the present invention are operable to generate a set of weights derived through crowdsourcing procedures for use in automatically performing white balancing operations on images captured by a digital camera system. Embodiments of the present invention are operable to generate a set of images which are illuminated with known and different illuminants. Using crowdsourcing procedures, embodiments of the present invention gather user feedback concerning which images from the set of images adjusted by the known illuminants are considered to be the most aesthetically pleasing. Images selected by the users are then stored within a database of selected images. Using a learning engine, embodiments of the present invention may then produce a set of weights based on the user selected images for use in determining a likely illuminant when performing automatic white balancing operations performed on the camera system.

Claims

1. A method of generating weights for use in a camera system, said method comprising:

generating an image set comprising a plurality of pairs of images wherein images of each pair are illuminated with known and different illuminants;

using crowdsourcing procedures, gathering user input from a plurality of users concerning a most pleasing image of each pair of images displayed to said plurality of users and storing said user input into a database of selected images; and

generating a set of weights for use in said camera system for selecting a proper illuminant for a newly captured image, said generating a set of weights performed using a learning engine operable to receive input from said database.

2. The method of generating weights for use in a camera system as described in claim 1, wherein said learning engine uses a support vector machine procedure.

3. The method of generating weights for use in a camera system as described in claim 1, wherein said gathering further comprises capturing a result of a user vote for said most pleasing image of each pair of images.

4. The method of generating weights for use in a camera system as described in claim 3, wherein said gathering further comprises screening said plurality of users for color selection competency.

5. The method of generating weights for use in a camera system as described in claim 1, wherein said gathering further comprises displaying said plurality of pairs of images to said users and wherein said displaying comprises randomizing a display ordering of said plurality of pairs of images displayed to said plurality of users.

6. The method of generating weights for use in a camera system as described in claim 1, wherein said generating a set of weights further comprises generating an updated set of weights responsive to new user input.

7. The method of generating weights for use in a camera system as described in claim 1, wherein said generating a set of weights further comprises receiving input from said database in a random order.

8. A system for generating weights for use in a camera system, said system comprising:

a generation module resident on a server system, wherein said generation module is operable to generate an image set comprising a plurality of pairs of images wherein images of each pair are illuminated with known and different illuminants;

a crowdsourcing module resident on said server system, wherein said crowdsourcing module is operable to gather user input from a plurality of users concerning a most pleasing image of each pair of images displayed to said plurality of users using crowdsourcing procedures and storing said user input into a database of selected images; and

a learning module resident on said server system, wherein said learning module is operable to process said database and responsive thereto produce therefrom a set of weights for use in said camera system for selecting a proper illuminant for a captured image.

9. The system for generating weights for use in a camera system as described in claim 8, wherein said learning module uses a support vector machine procedure.

10. The system for generating weights for use in a camera system as described in claim 8, wherein said crowdsourcing module is further operable to capture votes from said plurality of users to determine said most pleasing image of each pair of images.

11. The system for generating weights for use in a camera system as described in claim 8, wherein said crowdsourcing module is further operable to screen said plurality of users for color selection competency.

12. The system for generating weights for use in a camera system as described in claim 8, wherein said using learning module is further operable to generate an updated set of weights responsive to updated user input.

13. The system for generating weights for use in a camera system as described in claim 8, wherein said crowdsourcing module is further operable to randomize a display ordering of said plurality of pairs of images prior to being displayed to said plurality of users.

14. The system for generating weights for use in a camera system as described in claim 8, wherein said learning module is operable to receive input from said database in a random order.

15. A method of generating weights for use in a camera system, said method comprising:

receiving a classified image set comprising a plurality of pairs of images wherein images of each pair are illuminated with known and different illuminants;

using a server system, gathering user input using crowdsourcing procedures from a plurality of users concerning a most pleasing image of each pair of images displayed to said plurality of users and storing said user input into a database of selected images; and

generating a set of weights for use in said camera system for selecting a proper illuminant for a newly captured image, said generating a set of weights performed using a learning engine operable to receive input from said database.

16. The method of generating weights for use in a camera system as described in claim 15, wherein said learning engine uses a support vector machine procedure.

17. The method of generating weights for use in a camera system as described in claim 15, wherein said gathering further comprises receiving input from user votes concerning said most pleasing image of each pair of images.

18. The method of generating weights for use in a camera system as described in claim 15, wherein said gathering further comprises screening said plurality of users for color selection competency.

19. The method of generating weights for use in a camera system as described in claim 15, wherein said gathering further comprises displaying said plurality of pairs of images to said users and wherein said displaying comprises randomizing a display ordering of said plurality of pairs of images prior to being displayed to said plurality of users.

20. The method of generating weights for use in a camera system as described in claim 15, wherein said generating a set of weights further comprises generating an updated set of weights responsive to updated user input.

21. The method of generating weights for use in a camera system as described in claim 15, wherein said generating a set of weights further comprises receiving input from said database in a random order.