Using Red-Eye to Improve Face Detection in Low Quality Video Images

This paper presents a method to improve face detection by locating eyes in an image using infrared (IR) light, based on reflections due to the red-eye effect. The developed method analyzes two consecutive frames where the first one is taken under regular illumination with no IR and the second one using IR. The location of the eyes and the face contour are computed from the IR images using a collection of image processing techniques such as non-linear image denoising and enhancement, histogram stretch, image thresholding and edge detection. The algorithm operates successfully on low quality and low illumination gray images under a variety of skin tones, eye colors, angles, and illuminations, for subjects with and without glasses