Method for detecting eye structure and its apparatus

• US 5,432,866 A
• Filed: 06/14/1993
• Issued: 07/11/1995
• Est. Priority Date: 06/12/1992
• Status: Expired due to Fees

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Sakamoto, Shizuo
• Primary Examiner(s): Mancuso, Joseph
• Application Number:
  US08/075,800
• Time in Patent Office: 757 Days
• Field of Search: 382/1, 382/2, 382/6, 382/23, 382/50, 382/48, 382/9
• US Class Current: 382/128
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
• CPC Class Codes: G06V 40/168 Feature extraction; Face representation

Abstract

The method for detecting an eye structure and its apparatus according to this invention examines an image having an eye and its vicinity of a human face photographed from the vertex to the jaws to detect an area positioned between a position where brightness changes from lightness to darkness and a position where brightness changes from lightness to darkness and then to lightness, and detects from the above area an area whose width is larger than the pupil'"'"'s width and where an average of brightness values in the above area is darkest as the upper eyelid area. And, the inner canthus, outer canthus and their adjacent upper eyelid line are detected by finding the darkest point on the above image to detect the inner canthus and outer canthus points and their adjacent dot groups which become candidates for the upper eyelid line, and by carrying out the logical OR between the above upper eyelid area and the above dot groups.

Claims

1. A method for detecting an upper eyelid area comprising steps of:

examining an image of a vicinity of a human face photographed from the vertex to the jaws of the human face, said image including an eye, a pupil, and an upper eyelid area,

measuring how brightness of the image changes with position as measured at successive positions in a given direction on the image,

detecting a first position of the successive positions where the brightness of the image has decreased,

detecting a second position of the successive positions where the brightness has decreased and then increased,

detecting an intermediate area positioned between said first and second positions,

measuring, for a plurality of sub-areas within said intermediate area, averages of values of the brightness of the image, and

detecting from said intermediate area one of said sub-areas whose width is larger than the pupil'"'"'s width and whose average of values of the brightness of the image in said area is lowest among said averages of values, as the upper eyelid area.

2. A method for detecting an inner canthus, outer canthus, and upper eyelid area, comprising the steps of:

examining an image of a vicinity of a human face photographed from the vertex to the jaws of the human face, said image including an eye, a pupil, and an upper eyelid area,

measuring how brightness of the image changes with position as measured at successive positions in a given direction on the image,

detecting a first position of the successive positions where the brightness of the image has decreased,

detecting a second position of the successive positions where the brightness has decreased and then increased,

detecting an intermediate area positioned between said first and second positions,

measuring, for a plurality of sub-areas within said intermediate area, averages of values of the brightness of the image,

detecting from said intermediate area one of said sub-areas whose width is larger than the pupil'"'"'s width and whose average of values of the brightness of the image in said area is lowest among said averages of values, as the upper eyelid area,

finding the darkest point on said image to detect the inner canthus and outer canthus points and their adjacent dot groups which become candidates for the upper eyelid line, and

carrying out a logical OR between said upper eyelid area and said dot groups to detect the inner canthus, outer canthus, and their adjacent upper eyelid line.

3. A method for detecting an inner canthus, outer canthus and upper eyelid line comprising the steps of:

examining an image of a vicinity of a human face photographed from the vertex to the jaws of the human face, said image including an eye, a pupil, and an upper eyelid area,

measuring how brightness of the image changes with position as measured at successive positions in a given direction on the image,

detecting a first position of the successive positions where the brightness of the image has decreased,

detecting a second position of the successive positions where the brightness has decreased and then increased,

detecting an intermediate area positioned between said first and second positions,

measuring, for a plurality of sub-areas within said intermediate area, averages of values of the brightness of the image,

detecting from said intermediate area one of said sub-areas whose width is larger than the pupil'"'"'s width and whose average of values of the brightness of the image in said area is lowest among said averages of values, as the upper eyelid area,

finding the darkest point on said image to detect the inner canthus and outer canthus points and a primary dot group in their vicinity which become a candidate for the upper eyelid line,

carrying out a logical OR between said upper eyelid area and said dot groups to detect the inner canthus, outer canthus and the primary upper eyelid line in their vicinity,

examining said image from the vertex, while measuring how brightness of the image changes with position, detecting a point where the brightness of the image decreases, and carrying out the logical OR between the change point and said upper eyelid area to detect a secondary dot group which becomes a candidate for the upper eyelid, and

calculating a distance between said primary upper eyelid line and said secondary candidate dot group in a vertex direction to detect a dot group with smallest distance as a secondary upper eyelid line.

4. An apparatus for detecting an eye structure comprising:

a primary area detecting means for examining an image of a vicinity of a human face photographed from the vertex to the jaws of the human face, said image including an eye, a pupil, and an upper eyelid area, and detecting a first position where brightness of the image decreases, and a second position where the brightness of the image decreases and then increases, as measured along a direction on the image,

an upper eyelid candidate area detecting means for detecting from said primary area, as an upper eyelid candidate area, an area whose width is larger than a certain set value,

a secondary area detecting means for measuring, for a plurality of sub-areas within said upper eyelid candidate area, averages of values of the brightness of the image, and detecting as the upper eyelid area an area from among said sub-areas whose average of values of the brightness of the image in said area is lowest among said averages of values.

5. An apparatus for detecting an eye structure according to claim 4, wherein said certain set value is set on the basis of the pupil'"'"'s width.

6. An apparatus for detecting an eye structure according to claim 4, wherein said primary area detecting means comprises:

a first differential operating means for firstly differentiating said image from the vertex to the jaws,

a second differential operating means for secondly differentiating said image from the vertex to the jaws,

a primary binary-coding means for binary-coding said firstly differentiated image signal,

a secondary binary-coding means for binary-coding said secondly differentiated image signal,

a logical negative operating means for obtaining a logical negative signal from said binary-coded firstly differentiated image signal, and

a logical OR operating means for carrying out a logical OR between said firstly differentiated image signal of the logical negative and the binary-coded secondly differentiated image signal.

7. An apparatus for detecting an eye structure comprising:

an upper eyelid area detecting means for obtaining an upper eyelid area detected signal by detecting an upper eyelid area from an image of a vicinity of a human face, said image including an eye and an upper eyelid area,

a primary candidate detecting means for obtaining a primary candidate point detected signal by detecting a primary candidate point which is a point where brightness of the image first decreases and then increases as the image is examined in a vertex direction,

a secondary candidate detecting means for obtaining a secondary candidate point detected signal by detecting a secondary candidate point which is a point where brightness of the image first decreases and then increases as the image is examined from a direction perpendicular to the vertex direction, and

a logical OR operating means for carrying out a logical OR operation among all three of said upper eyelid area detected signal, said primary candidate point detected signal, and said secondary candidate point detected signal, to output as the inner canthus, outer canthus, and upper eyelid line detected signals.

8. An apparatus for detecting an eye structure according to claim 7, wherein the upper eyelid area detecting means comprises:

a primary area detecting means for detecting a first position where brightness changes from lightness to darkness, a second position where brightness changes from lightness to darkness and then to lightness, and an intermediate area positioned between the first and second positions, by examining said image, said image being of a vicinity of a human face photographed from the vertex to the jaws of the human face,

an upper eyelid candidate area detecting means for detecting from said primary area, as an upper eyelid candidate area, an area whose width is larger than a certain set value,

a secondary area detecting means for measuring, for a plurality of sub-areas within said upper eyelid candidate area, averages of values of the brightness of the image, and detecting as the upper eyelid area an area from among said sub-areas whose average of values of the brightness of the image in said area is lowest among said averages of values.

9. An apparatus for detecting an eye structure according to claim 7, wherein said primary candidate detecting means and secondary candidate detecting means comprise:

a first differential operating means for firstly differentiating said image in one of the vertex direction and the direction perpendicular to the vertex direction,

a second differential operating means for secondly differentiating said image in one of the vertex direction and the direction perpendicular to the vertex direction,

a binary-coding means for binary-coding said firstly differentiated image signal, and

a judging means for examining a code of said secondly differentiated image signal at the boundary of said binary-coded firstly differentiated image signal to output as the primary or secondary candidate point when the code is positive.

10. An apparatus for detecting an eye structure comprising:

an upper eyelid area detecting means for obtaining an upper eyelid area detected signal by detecting an upper eyelid area from an image of a vicinity of a human face photographed from the vertex to the jaws of the human face, said image including an eye, a pupil, and an upper eyelid area,

an inner canthus, outer canthus and upper eyelid line detecting means for detecting inner and outer canthus points and a primary upper eyelid line from said image,

a primary detecting means for obtaining primary point detected signal by detecting a primary point where brightness changes from lightness to darkness by examining said image from the vertex,

a secondary detecting means for detecting a candidate dot group which is a candidate for the upper eyelid area by carrying out a logical OR between the detected signal of said upper eyelid area and the detected signal of said primary point, and

an upper eyelid detecting means for detecting, from among a plurality of said candidate dot groups, a dot group as a secondary upper eyelid line by calculating, for each said candidate dot group, a distance between said inner and outer canthus points and primary upper eyelid line and said candidate dot group for the upper eyelid area in a vertex direction, and choosing the dot group that results in the smallest distance calculated.

11. An apparatus for detecting an eye structure according to claim 10, wherein said primary detecting means comprises:

a first differential operating means for firstly differentiating said image in one of the vertex direction and a direction perpendicular to the vertex direction,

a second differential operating means for secondly differentiating said image in one of the vertex direction and the direction perpendicular to the vertex direction,

a binary-coding means for binary-coding said secondly differentiated image signal, and

a judging means for examining a code of said firstly differentiated image signal at the boundary of said binary-coded secondly differentiated image signal to output as the primary point when it is positive.