Techniques for enabling or establishing the use of face recognition algorithms

• US 8,630,493 B2
• Filed: 12/07/2010
• Issued: 01/14/2014
• Est. Priority Date: 03/12/2006
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Gokturk, Salih Burak, Anguelov, Dragomir, Torresani, Lorenzo, Vanhoucke, Vincent, Shah, Munjal, Vu, Diem, Lee, Kuang-Chih
• Primary Examiner(s): Newman, Michael A
• Application Number:
  US12/962,494
• Publication Number:
  US 20110075934A1
• Time in Patent Office: 1,134 Days
• Field of Search: 382/203, 382/103, 382/115, 382/181, 382/118
• US Class Current: 382/203
• CPC Class Codes: G06V 10/7557 based on appearance, e.g. active appearance models [AAM]
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

Embodiments described herein facilitate or enhance the implementation of image recognition processes which can perform recognition on images to identify objects and/or faces by class or by people.

Claims

1. A method for performing recognition, method being implemented by one or more processors and comprising:

determining a region of an image where an object is likely located, the object of the image including a plurality of features;

identifying multiple sets of hypotheses locations using a plurality of probabilistic models for each of the plurality of features of the object, wherein each set of hypotheses locations corresponds to a plurality of possible locations of one of the plurality of features of the object;

selecting a most likely hypothesis location for each of the multiple sets of hypotheses locations based on the probabilistic model that provides the highest likelihood for the plurality of features of the object;

calculating one or more pixel values at the most likely hypothesis location of each set of hypotheses locations; and

locating, using the one or more pixel values, one or more pixels of the image that correspond to each of the plurality of features of the object.

2. The method of claim 1, wherein the object is a face, and wherein the plurality of features include one or more features selected from a set consisting of eyes, a nose tip, ends of a mouth, ear locations, hairline, and eyebrows, as well as other parts of the face.

3. The method of claim 1, wherein the plurality of features are based on a predefined model for a class of the object.

4. The method of claim 3, wherein the predefined model is a generic two-dimensional model.

5. The method of claim 1, wherein determining multiple sets of hypotheses locations includes deriving a set of hypotheses locations from a predefined object model.

6. The method of claim 1, further comprising determining a pose of the object.

7. The method of claim 3, wherein the predefined model is a generic three-dimensional model.

8. A method for performing recognition, method being implemented by one or more processors and comprising:

analyzing each image in a plurality of images, the plurality of images collectively depicting a plurality of different objects of an object class from various viewpoints or with various distortions, the plurality of images including a first image, depicting a first object of the object class with a first viewpoint or distortion, and a second image, depicting a second object of the object class with a second viewpoint or distortion that is different than the first viewpoint or distortion;

wherein analyzing each image in the plurality of images includes, for each image in the plurality of images;

determining a region of that image where an object of the object class is likely located, the object including a plurality of features,

identifying multiple sets of hypotheses locations using a plurality of probabilistic models for each of the plurality of features of the object, wherein each set of hypotheses locations corresponds to a plurality of possible locations of one of the plurality of features of the object depicted in the image,

selecting a most likely hypothesis location for each of the multiple sets of hypotheses locations based on the probabilistic model that provides the highest likelihood for the plurality of features of the object,

calculating one or more pixel values at the most likely hypothesis location of each set of hypotheses locations,

locating, using the one or more pixel values, one or more pixels of the image that correspond to each of the plurality of features of the object, and

wherein the method further comprises enabling at least the first object and the second object to be compared based on pixels of the first image that correspond to each of the plurality of features of the first object, and pixels of the second image that correspond to each of the plurality of features of the second object.

9. The method of claim 8, wherein the object class is a face.

10. The method of claim 9, wherein the first viewpoint or distortion of the object of the first image corresponds to a first facial expression, and wherein the second viewpoint or distortion of the object of the second image corresponds to a second facial expression.

11. The method of claim 8, wherein the plurality of features for the object depicted in each of the plurality of images include one or more features selected from a set consisting of eyes, a nose tip, ends of a mouth, ear locations, hairline, and eyebrows, as well as other parts of the face.

12. The method of claim 8, wherein the plurality of features for the object class are selected based on a predefined model for the object class.

13. The method of claim 12, wherein the predefined model is a generic two-dimensional model.

14. The method of claim 12, wherein the predefined model is a generic three dimensional generic model.

15. The method of claim 8, wherein determining the multiple sets of hypotheses locations includes deriving one or more of the multiple sets of hypotheses locations from a predefined object model.

16. The method of claim 8, further comprising determining a pose of the object depicted in one or more of the plurality of images.

17. The method of claim 8, further comprising comparing the first object of the first image to the second object of the same image to determine whether the first and second objects are the same object or of the same object class.

18. A computer system comprising:

a memory that stores instructions;

one or more processors, which access instructions from the memory to perform operations comprising;

determine a region of an image where an object is likely located, the object of the image including a plurality of features;

identify multiple sets of hypotheses locations using a plurality of probabilistic models for each of the plurality of features of the object, wherein each set of hypotheses locations corresponds to a plurality of possible locations of one of the plurality of features of the object;

select a most likely hypothesis location for each of the multiple sets of hypotheses locations based on the probabilistic model that provides the highest likelihood for the plurality of features of the object;

calculate one or more pixel values at the most likely hypothesis location of each set of hypotheses locations; and

locate, using the one or more pixel values, one or more pixels of the image that correspond to each of the plurality of features of the object.