Image processing apparatus and method, red-eye detection method, as well as programs for executing the image processing method and the red-eye detection method

• US 20050074164A1
• Filed: 09/17/2004
• Published: 04/07/2005
• Est. Priority Date: 09/19/2003
• Status: Active Grant

• Current Assignee: Fujifilm Corporation (Fujifilm Holdings Corporation)
• Original Assignee: Fuji Photo Film Co Limited (Fujifilm Holdings Corporation)
• Inventors: Yonaha, Makoto
• Granted Patent: US 7,450,739 B2
• Time in Patent Office: Days
• US Class Current: 382/162
• CPC Class Codes: G06F 18/00 Pattern recognition

Abstract

An image processing apparatus which processes an input image to detect a specified object of interest that is recorded in the input image is provided. The apparatus has a detection unit and a position determining unit. The detection unit has a plurality of identifying subunits each of which identifies an object from the input image under a different condition. The detection unit detects a region of the identified object. The position determining unit calculates the position coordinates of the object of interest in the input image from the region of the identified object. The detection unit performs the identification and the detection repeatedly and the detection unit performs the identification and the detection by using a detected region of an identified object in the previous identification. The detection unit detects the specified object of interest using the repeatedly detected regions.

Claims

1. An image processing apparatus which processes an input image to detect a specified object of interest that is recorded in the input image, said apparatus comprising:

a detection unit having a plurality of identifying subunits each of which identifies an object from the input image that satisfies a condition which is set differently each other, said detection unit detecting a region of the identified object; and

a position determining unit which calculates the position coordinates of said object of interest in the input image from the region of the identified object,

wherein said detection unit performs the identification and the detection repeatedly and, when the identification and the detection are repeated, said detection unit performs the identification and the detection under a different condition for identification than that used in previous identification of repeated identifications by using a detected region of an identified object in the previous identification, and said detection unit detects the specified object of interest using the repeatedly detected objects.

2. The image processing apparatus according to claim 1, wherein said detection unit has an identifying subunit that detects a resemblance to said object of interest as an object that is to be detected erroneously as said object of interest, and wherein a region of the detected resemblance is excluded from a region of said object of interest.

3. The image processing apparatus according to claim 1, wherein said input image is an image recorded with an image pickup device and said object of interest is an eye with the red-eye effect of a human subject or an animal.

4. The image processing apparatus according to claim 2, wherein said input image is an image recorded with an image pickup device, said object of interest is an eye with the red-eye effect of a human subject, and the identifying subunit for detecting said resemblance as an object is at least one member of the group consisting of an identifying subunit which detects spectacles worn by a human subject as the resemblance, an identifying subunit which detects the crow'"'"'s foot on the face as the resemblance, an identifying subunit which detects the eyebrow on the face as the resemblance, and an identifying subunit which detects a pimple on the face as the resemblance.

5. The image processing apparatus according to claim 1, wherein the identification of an object by said detection unit is such that the region of the object detected is narrowed down as the identification is repeated.

6. An image processing method which processes an input image to detect a specified object of interest that is recorded on the input image, said method comprising the steps of:

identifying an object from the input image that satisfy a condition;

detecting a region of the identified object;

repeating the step of identification and the step of detection a plurality of times with said condition being varied, thereby identifying and detecting an object each time under a different condition for than that used in previous identification of repeated identifications by using a detected region of an identified object in the previous identification; and

detecting a region of the specified object of interest within said input image on a basis of repeatedly detected regions of the identified objects and calculating the position coordinates of said object of interest in the input image from the detected region of the object of interest.

7. A red-eye detection method by which an input image that records a human subject or an animal is processed for detecting an eye with the red-eye effect of the human subject or the animal that is recorded in the input image, said method comprising the steps of:

detecting from the input image a possible region for red-eye that may have developed the red-eye effect;

detecting a region of a face of the human subject or the animal from said input image after detecting the possible region for red-eye; and

detecting the region of the eye with the red-eye effect from the input image by collating the results of detecting the possible region for red-eye and the face region.

8. The red-eye detection method according to claim 7, wherein at least one object selected from the group consisting of spectacles worn by a human subject, a crow'"'"'s foot on the human subject'"'"'s face, an eyebrow on the human subject'"'"'s face and a pimple on the human subject'"'"'s face is detected from said input image and the region of the eye with the red-eye effect is detected from said input image using the result of detection of at least one member of the group.

9. An executable program which causes a computer to perform image processing of an input image for detecting a specified object of interest that is recorded in the input image, said program comprising the steps of:

causing an operating unit in the computer to identify an object that satisfies a condition from within the input image;

causing said operating unit to detect a region of the identified object and storing it in a memory;

repeating the step of the identification and the step of the detection with said condition being varied, thereby causing said operating unit to perform a plurality of identification and detection such that a detected region of an identified object in previous identification of repeated identifications is used to get said operating unit to perform identification and detection under a different condition for identification than that used in the previous identification; and

causing said operating unit to detect a region of the object of interest within said input image on a basis of the regions of the repeatedly detected objects and causing said operating unit to calculate the position coordinates of said object of interest in the input image from the detected region of the object of interest.

10. An executable program that causes a computer to perform red-eye detection such that an input image that records a human subject or an animal is processed for detecting an eye with the red-eye effect of the human subject or animal that is recorded in the input image, said program comprising the steps of:

causing an operating unit in the computer to detect from the input image a possible region for red-eye that may have developed the red-eye effect;

causing said operating unit to function such that, after detecting the possible region for red-eye, said operating unit detects a region of a face of the human subject or animal from the input image using the result of detecting the possible region for red-eye; and

causing said operating unit to detect a region of the eye with the red-eye effect from the input image by collating results of the detections of the possible region for red-eye and the face region.

11. The program according to claim 10, which further includes at least one step selected from the group consisting of the step of causing said operating unit to function such that the spectacles worn by the human subject is detected from the input image as an object, the step of causing said operating unit to function such that a crow'"'"'s foot on the human subject'"'"'s face is detected from the input image as an object, the step of causing said operating unit to function such that the eyebrow on the human subject'"'"'s face is detected from the input image as an object, and the step of causing said operating unit to function such that a pimple on the human subject'"'"'s face is detected from the input image as an object,

and wherein the step by which the region of the eye with the red-eye effect is detected from the input image is assigned to said operating unit such that said operating unit detects the region of said eye with the red-eye effect using the result of at least one detecting step of the group which is followed by the step detecting the region of said eye with the red-eye effect.