摘要
为了能更加有效快速地进行眼睛定位,提出了一种基于块(based on blocks)的眼睛定位新算法,该算法首先将图像二值化并划分为块,然后利用两只眼睛的相似性和眼睛对的唯一性将一系列的图像块进行匹配,并以此确定一对眼睛的位置。由于该新算法利用了两眼睛固有的位置关系与相互间的相似性,将两眼睛成对考虑,所以能检测各种角度(平面旋转)人脸的眼睛,并可得到较高的检测准确率。此外,实验显示该算法在一定程度上还能适应不同光照条件和表情变化。大量人脸图片上的实验结果表明,该算法可以可靠?快速地定位眼睛。
In this paper, we introduce an effective algorithm based on blocks （BoB） for eye location. The image is partitioned to blocks by a threshold. Based on the comparability between two eyes and the uniqueness of a pair of eyes, it can locate eyes after matching each two blocks. Due to what we use are the location and the comparability between two eyes and eyes are considered in pairs, the algorithm could work on rotated images and get higher detection rate. What＇s more, it can deal with moderate illumination and expression changes. Experiments on a number of face images have shown that our algorithm is reliable and fast.
出处
《中国图象图形学报》
CSCD
北大核心
2007年第10期1841-1844,共4页
Journal of Image and Graphics
关键词
面部特征
眼睛定位
人脸
识别块
facial feature, eye localization, face recognition, block
