Human eye detection method, apparatus, system and storage medium

• US 7,620,214 B2
• Filed: 12/17/2003
• Issued: 11/17/2009
• Est. Priority Date: 12/31/2002
• Status: Expired due to Fees

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Wang, Libing, Chen, Xinwu, Ji, Xin
• Primary Examiner(s): Bella; Matthew C
• Assistant Examiner(s): Newman; Michael A
• Application Number:
  US10/736,686
• Publication Number:
  US 20040136592A1
• Time in Patent Office: 2,162 Days
• Field of Search: 348/78, 382/103, 382/104, 382/117, 382/194, 382/115, 382/118, 382/206, 351/201, 351/208, 351/204, 351/203
• US Class Current: 382/117
• CPC Class Codes: G06V 40/18 Eye characteristics, e.g. of the iris

Abstract

The invention relates to a human eye detection method, apparatus, system and storage medium. According to the invention, candidate eye areas are detected from a given image by: reading in the image; analyzing the image and getting a list of candidate eye areas; selecting one unverified candidate eye area from said list; determining a neighborhood region of the selected candidate eye area; calculating the neighborhood region'"'"'s size, which is recorded as S; processing the region, obtaining dark areas; counting the number of dark areas, which number is recorded as N; comparing the ratio N/S to a predetermined first threshold, if the ratio N/S is smaller than said first threshold, the candidate eye area is determined as a false eye area and deleted from the list; else, the candidate eye area is determined as a real eye area and remains in the list; repeating foresaid selecting step to comparing step until there is no unverified candidate eye area in the list; and outputting the list for purpose of subsequent processing of the image. The invention may be applied to human body detection.

Claims

1. A human eye detection method for detecting a human eye by determining whether an unverified candidate eye area is a true eye area based on a neighborhood region of the unverified candidate eye area, comprising the following steps:

a) inputting an image into an information processing apparatus;

b) analyzing the image and getting the unverified candidate eye area, in the information processing apparatus;

c) determining, in the information processing apparatus, the neighborhood region based on the unverified candidate eye area, the neighborhood region being defined as rectangle, the center of which is the center of the unverified candidate eye area and the size of which is calculated from the size of the image;

d) calculating, in the information processing apparatus, the neighborhood region'"'"'s size, which is recorded as S;

e) detecting, in the information processing apparatus, dark areas in the neighborhood region and counting the number of dark areas, which number is recorded as N; and

f) determining, in the information processing apparatus, whether or not the unverified candidate eye area is a true eye area by comparing the ratio N/S to a predetermined threshold, wherein if the ratio N/S is smaller than said threshold, the unverified candidate eye area is determined to be a true eye area;

else, the unverified candidate eye area is determined to be a false eye area.

2. The human eye detection method according to claim 1, characterized in that the method further comprises the step of:

determining candidate face areas on the basis of a candidate eye area determined to be a true eye area in step f).

3. The human eye detection method according to claim 1 or 2, characterized in that, step e), executes a binarization processing for detecting the dark areas.

4. The human eye detection method according to claim 1 or 2, further comprising a threshold calculating step before the step f), for calculating said threshold.

5. A human eye detection apparatus for detecting a human eye by determining whether an unverified candidate eye area is a true eye area based on a neighborhood region of the unverified candidate eye area, comprising:

an input unit that inputs an image; and

a processor that (i) analyzes the image to obtain the unverified candidate eye area;

(ii) determines the neighborhood region based on the unverified candidate eye area, the neighborhood region being defined as a rectangle, the center of which is the center of the unverified candidate eye area and the size of which is calculated from the size of the image, (iii) calculates the neighborhood region'"'"'s size S, (iv) detects dark areas in the neighborhood region and determines the total count N of dark areas in the neighborhood region, (v) and determines whether or not the unverified candidate eye area is a true eye area by comparing the ratio N/S to a predetermined threshold, wherein if the ratio N/S is smaller than the threshold, the unverified candidate eye area is determined to be a true eye area, else the unverified candidate eye area is determined to be a false eye area.

6. A human eye detection apparatus according to claim 5, wherein said processor executes a binarization processing to detect the dark areas.

7. A computer-readable storage medium embodying program codes for causing an apparatus to perform a human eye detection method for detecting a human eye by determining whether an unverified candidate eye area is a true eye area based on a neighborhood region of the unverified candidate eye area, comprising:

inputting an image;

analyzing the image and getting the unverified candidate eye area;

determining the neighborhood region based on the unverified candidate eye area, the neighborhood region being defined as a rectangle, the center of which is the center of the unverified candidate eye area and the size of which is calculated from the size of the image;

calculating the neighborhood region'"'"'s size, which is recorded as S;

detecting dark areas in the neighborhood region and counting the number of dark areas, which number is recorded as N; and

determining whether or not the unverified candidate eye area is a true eye area by comparing the ratio N/S to a predetermined first threshold, wherein if the ratio N/S is smaller than the threshold, the unverified candidate eye area is determined to be a true eye area, else the unverified candidate eye area is determined to be a false eye area.