一种基于准二值图的眼仁定位新方法  

A Human Eye Location Algorithm Based on Quasi Binary Image

作　　者：高岩[1] 赵丽娜[1] 

机构地区：[1]北京化工大学理学院,北京100029

出　　处：《吉林大学学报（理学版）》2009年第5期999-1005,共7页Journal of Jilin University:Science Edition

基　　金：国家自然科学基金(批准号:60803099)

摘　　要：提出一种基于以二次二值化图像为研究对象,二次垂直投影为定位手段的眼睛定位方法.先用小波8层分解原图像,垂直和水平投影定出眼部区域的坐标;再对原图像进行奇异值分解,取最大奇异值的1/3.5为新值,对图形进行重建得到对眼仁有较好刻画的准二值图.用已知的眼部坐标在无噪声准黑白图上截取眼睛部分后,再次用垂直投影方法得到眼仁的准确位置.A human eye localization method is described in this paper. It＇ s based on twice binary images and twice vertical projection. Firstly, the original image is decomposed with 8-level of wavelet, and then the main area for eyes is gotten by vertical and horizontal projection. Secondly, the given image is decomposed with SVD again and a max singular value of 1/3.5 is set to be a new threshold and to reconstruct the original image for obtaining the quasi binary image which embodies eyes better. At last, eyes area is intercepted from less-noise quasi binary image by already known coordinates and finally, the accurate position of eyes is achieved by vertical projection once again.

关 键 词：奇异值分解 小波分解 投影函数 人眼定位 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]