Eyes Detection Based on One-Dimensional Wavelet Transform

This paper presents an algorithm for eyes detection based on one-dimensional wavelet transform and gray-level integration projection. First, nonlinear color transformation is used to detect skin color region, which can be used to acquired possible skin areas. According to the geometric characteristics of eyes, horizontal integral projection is used to upper face, then using wavelet transform for projecting curve in order to exclude interference, and eyes approximate area can be extracted. Finally, vertical integral projection is used to eyes approximate position and wavelet transform is used for projection curve, thus exact positions of eyes can be located. This paper uses the camera to capture the face image, and the algorithm has advantages of low computation complexity, high speed and better detection effect.