Image analysis – Applications – Personnel identification
Reexamination Certificate
1998-03-24
2001-04-10
Mehta, Bhavesh (Department: 2621)
Image analysis
Applications
Personnel identification
C348S078000, C351S206000
Reexamination Certificate
active
06215891
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an eye image recognition method for extracting a region of a pupil of an eye from a captured image including the eye, and further relates to eye image selection method and system for selecting an image including eye data about a pupil and its peripheral portion of an eye from among consecutively captured images each including the eye. The eye data is used for individual identification of an animal, such as a human being, a horse or a cow.
2. Description of the Related Art
A technique is disclosed, for example, in JP-A-4-264985, for carrying out individual identification based on a human face. In the disclosed technique, regions of the eyes, nose and so on are individually extracted from an input image of the whole face, captured by a camera, so as to be subjected to identification processes, respectively. For locating, for example, the eye region to be extracted, a rectangular region is preset in the input face image and projected in vertical and horizontal directions, and positions having the minimum luminance value in the projected regions are considered to define the center of the eye.
In the foregoing technique, it is necessary that the human face is adequately positioned relative to the camera so that the eyes and so forth can be present in the corresponding preset rectangular regions.
However, in practice, human faces are not always positioned adequately relative to cameras, and further, it is almost impossible to adequately position faces of other animals relative to cameras. Thus, for automatically searching out the position of an eye for extraction, it is necessary to select an image, which includes the eye in a preset region, from a number of consecutively captured images.
In view of this, a technique has been demanded which can achieve automatic selection of such an image from the consecutively captured images.
On the other hand, when carrying out individual identification using eye data of a human being or an animal of another kind, it is necessary to use an enlarged image of an eye for obtaining as many eye data as possible. However, in the enlarged eye image, those portions, such as an eyebrow, eyelashes, a mole and a shadow, may appear as regions having the same or like density (luminance) values as a pupil of the eye. Therefore, it is difficult to identify a region of the pupil only based on density data, such as differences in density.
Specifically, as shown in
FIG. 2
, an eyebrow, eyelashes and a mole exist in an image of an eye as regions having the same or like density values as the pupil of the eye. Accordingly, when the image is simply projected in vertical and horizontal directions, a plurality of minimum value regions are located so that it is difficult to extract only the region of the pupil with accuracy.
In view of this, a technique has been demanded which can extract the pupil region with accuracy even if the input image includes portions having the same or like density values as the pupil region.
SUMMARY OF THE INVENTION
Therefore, it is an object of the present invention to provide an improved eye image recognition method.
It is another object of the present invention to provide an improved eye image selection method.
It is still another object of the present invention to provide an improved eye image selection system.
According to one aspect of the present invention, an eye image recognition method comprises the steps of dividing an input image including an eye into a plurality of blocks; producing a mosaic image with the blocks each representing a density value; and determining that one of the blocks having a small distance from a center point of the mosaic image and having a small density value is a block including a center position of a pupil of the eye.
It may be arranged that the method further comprises the steps of setting a threshold value based on the density value of the determined block; converting the input image into a two-valued image using the threshold value as a criterion; and extracting, as a pupil region, a region of the two-valued image including the center position of the pupil.
According to another aspect of the present invention, an eye image selection method comprises the steps of locating a center position of a pupil of an eye from an input image including the eye; and determining that, if a distance between the center position of the pupil and a center position of the input image is not greater than a preset value, the input image includes necessary eye data for carrying out individual identification, and selecting the input image.
It may be arranged that the method further comprises the steps of dividing the input image into a plurality of blocks; producing a mosaic image with the blocks each representing a density value; determining one of the blocks having a small distance from a center point of the mosaic image and having a small density value; and locating the center position of the pupil based on the determined block.
According to another aspect of the present invention, an eye image selection method comprises the steps of locating a center position of a pupil of an eye from an input image including the eye; and determining that, if a density of a region including the center position is not greater than a given value, the input image includes the pupil necessary for carrying out individual identification, and selecting the input image.
It may be arranged that the method further comprises the steps of dividing the input image into a plurality of blocks; producing a mosaic image with the blocks each representing a density value; determining one of the blocks having a small distance from a center point of the mosaic image and having a small density value; and locating the center position of the pupil based on the determined block.
According to another aspect of the present invention, an eye image selection method comprises the steps of dividing an input image including an eye into a plurality of blocks; producing a mosaic image with the blocks each representing a density value; determining that one of the blocks having a small distance from a center point of the mosaic image and having a small density value is a block including a center position of a pupil of the eye; locating the center position of the pupil from the determined block; and selecting the input image if a distance between the center position of the pupil and a center position of the input image is not greater than a preset value and if a density of a region including the center position of the pupil is not greater than a preset value.
According to another aspect of the present invention, an eye image selecting method of selecting an input image for individual identification from a plurality of input images each including an eye, the method comprises the steps of locating, in the input image, a region of a specular reflection image appearing in the eye due to a light source upon capturing the input image; and selecting the input image when the region satisfies a preset condition.
It may be arranged that the preset condition is whether the region exists in a preset range of the input image.
It may be arranged that the preset condition is whether a size of the region is in a preset range.
It may be arranged that the preset condition includes an approximation of a shape of the region relative to a shape of the light source.
It may be arranged that the preset condition includes an edge intensity of a contour of the region relative to a threshold value.
It may be arranged that the preset condition is whether a density value of a region around the region of the specular reflection image is smaller than a threshold value.
It may be arranged that the region of the specular reflection image is located by locating regions having density values greater than a threshold value in the input image and by selecting one of the located regions as appearing on a cornea of the eye when none of the other located regions exist in the neighborhood of the one of the located regions.
It may
Kuno Yuji
Suzaki Masahiko
Frank Robert J.
Mehta Bhavesh
OKI Electric Industry Co., Ltd.
Venable
Wood Allen
LandOfFree
Eye image recognition method eye image selection method and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Eye image recognition method eye image selection method and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Eye image recognition method eye image selection method and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2536911