| 摘要 |
PROBLEM TO BE SOLVED: To automatically detect eye positions using a simple technique from an original image that includes a plurality of face images within one image. SOLUTION: A method for detecting eye positions from the original image comprises: a process A including (A1) creating a blurred image obtained by blurring of the original image in two intersecting directions, and determining the position of an intersection of edges in the blurring directions, (A2) creating an image for face area detection where the gradation of the original image is broken, and (A3) in the image for face area detection, detecting face area candidates from a rectangular area one of the apexes of which is the intersection of the edges in the blurring directions; and a process B including (B1) creating a plurality of images for eye position detection with varying brightness for each of the face area candidates, (B2) as the images for eye position detection fade in from a faded-out condition, detecting the block areas of pixels that appear gradually in the images, (B3) selecting those of the detected block areas of pixels that have appeared in pairs as eye position candidates, and (B4) specifying the eye position candidates as eye positions based on their frequencies of appearance. COPYRIGHT: (C)2009,JPO&INPIT
|
