Image analysis – Applications – Personnel identification
Reexamination Certificate
2001-02-22
2003-06-03
Do, Anh Hong (Department: 2624)
Image analysis
Applications
Personnel identification
C382S170000, C382S224000
Reexamination Certificate
active
06574354
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to the field of image detection and more specifically to the detection of a face disposed within a digital image.
As broadcasting becomes more digitally based, it becomes easier to archive and catalog video content. Researchers have developed systems for content-based image and video indexing and retrieval that utilize low-level visual features (semantics) like color, texture, shape and sketch of an image. To facilitate the automatic archiving and retrieval of video material based on higher level semantics, it is important to detect and recognize events in video clips. Human activities are important events in video clips and face detection is a step toward the recognization of human activities.
Face detection is also useful in security systems, criminal identifications, digital image capturing, and teleconferences. In a security system, for example, it is useful to detect the facial portions of an image being viewed so that an operator of the system can discern whether a human is present in the image.
The detection of faces from images has not received much attention by researchers. Most conventional techniques concentrate on face recognition and assume that a face has been identified within the image or assume that the image only has one face as in a “mug shot” image. Such conventional techniques are unable to detect faces from complex backgrounds.
One prior art technique that does perform face detection determines whether a cluster of pixels conforms to a facial template. This routine is deficient because of different scales and orientations of possible faces. The template itself is one size and orientation and will not detect faces which are of a different size or are rotated. Consequently, the template itself must be scaled up and down and rotated while searching is performed yielding a search space that is too big to be useful or practical. Some prior art techniques, like EPA 0836326 A2, use merely a shape template to see if a cluster of pixels conforms to that shape. In addition to the scaling and rotation problems mentioned above, this solution is too simplistic to be used with complex backgrounds which may have many objects with the same shape as a face and perhaps even the same color as a face.
Therefore, there exists a need for a method of detecting faces within a digital image in which the face is disposed within a complex background.
SUMMARY AND OBJECTS OF THE INVENTION
One aspect of the invention is a method for detecting a face disposed within a digital image. The method comprises providing a digital image composed of a plurality of pixels and producing a binary image from the digital image by detecting skin colored pixels. The method further includes removing pixels corresponding to edges in the luminance component thereby producing binary image components; mapping the binary image components into at least one graph; and classifying the mapped binary image components as facial and non-facial types wherein the facial types serve as facial candidates.
In this way, as well as with the following aspects of the invention, a face disposed within a digital image can be detected.
Another aspect of the invention is a computer readable storage medium containing data for performing the steps of providing a digital image composed of a plurality of pixels and producing a binary image from the digital image by detecting skin colored pixels. The computer readable storage medium further contains data for removing pixels corresponding to edges in the luminance component thereby producing binary image components; mapping the binary image components into at least one graph; and classifying the mapped binary image components as facial and non-facial types wherein the facial types serve as facial candidates.
Yet another aspect of the invention is a method for detecting a face disposed within a digital image. The method comprises providing a digital image composed of a plurality of pixels and producing a binary image from the digital image by detecting skin colored pixels. The method further includes removing pixels corresponding to edges in the luminance component thereby producing binary image components; and classifying each of the binary image components as one of a facial type and a non-facial type. The classifying includes forming a bounding box around a classified component of the components and performing at least one of: comparing an area of the bounding box to a bounding box threshold; comparing an aspect ratio of the bounding box to an aspect ratio threshold, determining an area ratio, the area ratio being the comparison between the area of the classified component and the area of the bounding box, and comparing the area ratio to an area ratio threshold; determining an orientation of elongated objects within the bounding box; and determining a distance between a center of the bounding box and a center of the classified component.
It is an object of the invention to provide a method for detecting a face disposed within a digital image.
This object, as well as others, will become more apparent from the following description read in conjunction with the accompanying drawings where like reference numerals are intended to designate the same elements.
REFERENCES:
patent: 6072903 (2000-06-01), Maki et al.
patent: 6263113 (2001-07-01), Abbdel-Mottaleb et al.
Takacs et al., “Face Recognition Using binary Image Metrics”, Third IEEE Interconference on Automastic Face and Gesture Recognition, 1998, pp. 294-299.*
Terrillon et al., “Automatic Detection of Human Faces in Natural Scene Images by Use of a Skin Color Model and of Invariant Moments”, Third IEEE Interconference on Automatic Face and Gesture Recognition, 1998, pp. 112-117.
Abdel-Mottaleb Mohamed
Elgammal Ahmed
Do Anh Hong
Koninklijke Philips Electronics , N.V.
LandOfFree
Method for detecting a face in a digital image does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for detecting a face in a digital image, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for detecting a face in a digital image will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3089141