System and method for face detection through geometric distribution of a non-intensity image property
First Claim
1. A system for detecting a face in an image, comprising:
- a feature extraction module that extracts features from the image using a non-intensity image property;
a feature template that determines a high-frequency variation around each pixel in the image;
a feature averaging module that groups the features into facial regions;
a relational template module that uses a relational template over a geometric distribution of the non-intensity image property; and
regional values assigned by the relational template module that are based on a relationship between the facial regions and are used to determine whether the face has been detected.
2 Assignments
0 Petitions
Accused Products
Abstract
The present invention is embodied in a system and method for detecting a face within an image using a relational template over a geometric distribution of a non-intensity image property. In general, the system of the present invention includes a hypothesis module for defining a sub-region in which to search for a face, a feature extraction module for extracting image feature values image based on a non-intensity image property, an averaging module for grouping the extracted image feature values into geometrically distributed facial regions, and a relational template module that uses a relational template and facial regions to determine whether a face has been detected. In a preferred embodiment the image property used is edge density, although other suitable properties (such as pixel color) may also be used. The method of the present invention includes performing feature extraction on the image based on an image property (such as edge density), grouping extracted image feature values into facial regions and using a relational template to determine whether a face has been detected.
167 Citations
39 Claims
-
1. A system for detecting a face in an image, comprising:
-
a feature extraction module that extracts features from the image using a non-intensity image property;
a feature template that determines a high-frequency variation around each pixel in the image;
a feature averaging module that groups the features into facial regions;
a relational template module that uses a relational template over a geometric distribution of the non-intensity image property; and
regional values assigned by the relational template module that are based on a relationship between the facial regions and are used to determine whether the face has been detected. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A method for detecting a face in an image, comprising:
-
extracting features from the image using a non-intensity image property;
determining a high-frequency variation around each pixel in the image;
grouping the extracted features into facial regions based on a geometric distribution of the non-intensity image property; and
determining whether a face has been detected in the image by using a relational template over a geometric distribution of the non-intensity property and a relationship between the facial regions. - View Dependent Claims (17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33)
generating a hypothesis about the location of the face within the image; and
defining a sub-region within the image to search for the face.
-
-
30. The method of claim 29, wherein the hypothesis is based on at least one of:
- (a) a scale;
(b) an aspect ratio;
(c) a location in the image.
- (a) a scale;
-
31. The method of claim 29, wherein the hypothesis is based at least partially on the non-intensity image property.
-
32. The method of claim 16, further comprising preprocessing the image.
-
33. The method of claim 16, wherein preprocessing comprises at least one of:
- (a) resizing the image;
(b) filtering out unwanted noise from the image;
(c) performing histogram equalization on the image.
- (a) resizing the image;
-
34. A method for detecting a face in an image, comprising:
-
extracting features from the image using a non-intensity image property and a feature template to determine high-frequency variation around each pixel of the image;
placing the features into facial regions;
using a relational template over a geometric distribution of the non-intensity property to determine a relationship between the facial regions;
defining a relational value based on the relational template; and
detecting a face within the image if the relational value is greater than a threshold value;
wherein the non-intensity image property includes at least one of;
(a) edge density;
(b) pixel color.- View Dependent Claims (35, 36, 37)
-
-
38. A computer-implemented method for deciding whether a face is present in a image, comprising:
-
generating an hypothesis as to a location in the image where a face may be present;
defining a sub-region in the image based on the generated hypothesis;
extracting image feature values based on a non-intensity image property for each pixel in the sub-region using a feature template that is sensitive to high frequencies to determine the degree of high-frequency variation around each pixel in the sub-region;
defining facial regions in the sub-region of the image;
combining each of the image feature values in each of the facial regions to obtain a combined feature value for each of the facial regions;
determining a relationship between any two of the facial regions and assigning a regional value based on the relationship;
summing the regional values to obtain a relational value; and
comparing the relational value to a threshold value to decide whether a face is present in the image. - View Dependent Claims (39)
-
Specification