Human eye feature extraction based on segmented binarization

For facial feature extraction problems, eye feature extraction is a basic and critical work. This paper presents a human eye feature extraction method based on segmented binarization. After binarization we detect the eye corner points by the corner detector which is based on the curve scale space. The detector uses adaptive threshold and dynamic region of support. Due to the corner detection errors, we propose a method which uses the mean of the local area around the corner to correct them. Experiments show that the proposed method is effective.