Apparatus and method for detecting speaking person's eyes and face

Abstract

An apparatus for detecting the position of a human face in an input image or video image and a method thereof are provided. The apparatus includes an eye position detecting means for detecting pixels having a strong gray characteristic to determine areas having locality and texture characteristics as eye candidate areas among areas formed by the detected pixels, in an input red, blue, and green (RGB) image, a face position determining means for creating search templates by matching a model template to two areas extracted from the eye candidate areas, and determining an optimum search template among the created search templates by using the value normalizing the sum of a probability distance for the chromaticity of pixels within the area of a search template, and horizontal edge sizes calculated in the positions of the left and right eyes, a mouth and a nose estimated by the search template, and an extraction position stabilizing means for forming a minimum boundary rectangle by the optimum search template, and increasing count values corresponding to the minimum boundary rectangle area and reducing count values corresponding to an area other than the minimum boundary rectangle area, among count values of individual pixels, stored in a shape memory, to output the area in which count values above a predetermined value are positioned, as eye and face areas. The apparatus is capable of accurately and quickly detecting a speaking person's eyes and face in an image, and is tolerant of image noise.

Dependencies are inferred from the claim text.