Distinguishing between faces and non-faces
Methods, systems, and apparatus, including computer program products, for evaluating image data. In one aspect, a method includes accessing an image that includes a candidate face, such as a face detected during a face detection operation. The method further includes generating a sharpness measure based on image data corresponding to the candidate face, evaluating the sharpness measure to determine a confidence score representing a likelihood that the candidate face corresponds to a human face, and accepting the candidate face when the confidence score compares in a predetermined manner to a confidence threshold. Additionally, the method can be implemented to include generating a skin tone measure based on image data corresponding to the candidate face and evaluating the sharpness measure in combination with the skin tone measure to determine the confidence score.
1. A computer-implemented method for evaluating image data, the method comprising:
accessing an image that includes a candidate face;
generating a sharpness measure based on image data corresponding to the candidate face;
evaluating the sharpness measure to determine a confidence score representing a likelihood that the candidate face corresponds to a human face; and
accepting the candidate face when the confidence score compares in a predetermined manner to a confidence threshold.
2. The computer-implemented method of claim 1 , wherein generating the sharpness measure further comprises:
generating, for an image portion associated with the candidate face, a total sharpness score representing a difference between a lateral sharpness value and a corresponding longitudinal sharpness value.
3. The computer-implemented method of claim 1 , further comprising:
generating the sharpness measure in accordance with data associated with a composite face.
4. The computer-implemented method of claim 3 , further comprising:
determining, for an image portion associated with the candidate face, a sharpness value across facial characteristics indicated by the composite face and a sharpness value along facial characteristics indicated by the composite face; and
determining a difference between the sharpness value across facial characteristics and the sharpness value along facial characteristics to generate the sharpness measure.
5. The computer-implemented method of claim 4 , further comprising:
generating, for the image portion, a second sharpness measure representing a difference between a lateral sharpness value and a corresponding longitudinal sharpness value; and
evaluating, in combination, the sharpness measure and the second sharpness measure to determine the confidence score.
6. The computer-implemented method of claim 1 , further comprising:
generating a skin tone measure based on image data corresponding to the candidate face; and
evaluating the sharpness measure in combination with the skin tone measure to determine the confidence score.
7. The computer-implemented method of claim 6 , wherein evaluating the sharpness measure in combination with the skin tone measure further comprises:
assigning a weighting factor to at least one of the sharpness measure and the skin tone measure.
8. The computer-implemented method of claim 1 , wherein the confidence threshold comprises a dynamic threshold.
9. A computer-implemented method for evaluating image data, the method comprising:
accessing an image that includes a candidate face;
scaling the candidate face in accordance with a predetermined standard size to generate a scaled candidate face;
generating a sharpness measure and a skin tone measure based on the scaled candidate face;
evaluating the sharpness measure and the skin tone measure to determine a confidence score; and
accepting the candidate face when the confidence score compares in a predetermined manner to a confidence threshold.
10. The computer-implemented method of claim 9 , further comprising:
identifying a direction of a facial characteristic associated with a portion of a composite face;
determining, for a corresponding image portion associated with the scaled candidate face, a sharpness value across the facial characteristic and a sharpness value along the facial characteristic; and
determining a difference between the sharpness value across the facial characteristic and the sharpness value along the facial characteristic to generate the sharpness measure.
11. The computer-implemented method of claim 10 , wherein the portion of the composite face corresponds to an eye region.
12. The computer-implemented method of claim 10 , further comprising:
generating, for the image portion, a second sharpness measure representing a difference between a lateral sharpness value and a corresponding longitudinal sharpness value; and
evaluating, in combination, the sharpness measure and the second sharpness measure to determine the confidence score.
13. The computer-implemented method of claim 12 , further comprising:
comparing the second sharpness measure with a corresponding sharpness measure for the composite face to determine a degree of similarity.
14. The computer-implemented method of claim 9 , further comprising:
assigning a weighting factor to at least one of the sharpness measure and the skin tone measure.
15. The computer-implemented method of claim 9 , further comprising:
generating a head-tilt angle based on the scaled candidate face; and
determine the confidence score in accordance with the sharpness measure, the skin tone measure, and the head-tilt angle.
16. A system comprising:
a computer-readable medium storing an image that includes a candidate face; and
a computing system including processor electronics configured to perform operations comprising:
accessing the candidate face;
generating a sharpness measure and a skin tone measure based on image data corresponding to the candidate face;
evaluating the sharpness measure and the skin tone measure to determine a confidence score representing a likelihood that the candidate face corresponds to a human face; and
accepting the candidate face when the confidence score compares in a predetermined manner to a confidence threshold.
17. The system of claim 16 , wherein the processor electronics are further configured to perform operations comprising:
assigning a weighting factor to at least one of the sharpness measure and the skin tone measure.
18. The system of claim 17 , wherein the processor electronics are further configured to perform operations comprising:
executing, by a face evaluation processor, a training procedure to generate one or more operating parameters; and
determining, by the face evaluation processor, the weighting factor in accordance with the one or more operating parameters.
19. A computer program product, encoded on a non-transitory computer-readable medium, operable to cause data processing apparatus to perform operations comprising:
accessing an image that includes a candidate face;
generating a sharpness measure based on image data corresponding to the candidate face;
evaluating the sharpness measure to determine a confidence score representing a likelihood that the candidate face corresponds to a human face; and
accepting the candidate face when the confidence score compares in a predetermined manner to a confidence threshold.
20. The computer program product of claim 19 , further operable to cause data processing apparatus to perform operations comprising:
generating, for an image portion associated with the candidate face, a total sharpness score representing a difference between a lateral sharpness value and a corresponding longitudinal sharpness value.
21. The computer program product of claim 19 , further operable to cause data processing apparatus to perform operations comprising:
identifying a direction of a facial characteristic associated with a portion of a composite face;
determining, for a corresponding image portion associated with the candidate face, a sharpness value across the facial characteristic and a sharpness value along the facial characteristic; and
determining a difference between the sharpness value across the facial characteristic and the sharpness value along the facial characteristic to generate the sharpness measure.
22. The computer program product of claim 21 , further operable to cause data processing apparatus to perform operations comprising:
generating, for the image portion, a second sharpness measure representing a difference between a lateral sharpness value and a corresponding longitudinal sharpness value; and
evaluating, in combination, the sharpness measure and the second sharpness measure to determine the confidence score.
23. The computer program product of claim 19 , further operable to cause data processing apparatus to perform operations comprising:
sampling pixels corresponding to the candidate face in accordance with a pattern; and
generating a skin tone measure based on the sampled pixels.
24. The computer program product of claim 23 , further operable to cause data processing apparatus to perform operations comprising:
evaluating the sharpness measure in combination with the skin tone measure to determine the confidence score.
25. The computer program product of claim 24 , further operable to cause data processing apparatus to perform operations comprising:
weighting at least one of the sharpness measure and the skin tone measure prior to determining the confidence score.
26. A computer program product, tangibly encoded on a computer-readable medium, operable to cause data processing apparatus to perform operations comprising:
accessing an image that includes a candidate face;
scaling the candidate face in accordance with a predetermined standard size to generate a scaled candidate face;
generating a sharpness measure and a skin tone measure based on the scaled candidate face;
evaluating the sharpness measure and the skin tone measure to determine a confidence score; and
accepting the candidate face when the confidence score compares in a predetermined manner to a confidence threshold.
27. The computer program product of claim 26 , further operable to cause data processing apparatus to perform operations comprising:
analyzing a portion of a composite face to identify a direction of a facial characteristic;
determining, for a corresponding image portion associated with the scaled candidate face, a sharpness value across the facial characteristic and a sharpness value along the facial characteristic; and
generating the sharpness measure based on a difference between the sharpness value across the facial characteristic and the sharpness value along the facial characteristic.
28. The computer program product of claim 27 , further operable to cause data processing apparatus to perform operations comprising:
selecting the portion of the composite face based on a level of image detail.
29. The computer program product of claim 26 , further operable to cause data processing apparatus to perform operations comprising:
selecting an image portion of the scaled candidate face;
determining a lateral sharpness value and a longitudinal sharpness value for the selected image portion; and
generating the sharpness measure based on the determined lateral sharpness value and the determined longitudinal sharpness value.
30. The computer program product of claim 26 , further operable to cause data processing apparatus to perform operations comprising:
sampling pixels corresponding to the candidate face; and
generating a skin tone measure based on the sampled pixels.