基于累积帧差的人眼定位及模板提取

提出了一种基于累积帧差的眼睛定位方法。利用VC 下的VFW软件包,建立一个对实时捕获的具有眨眼状态的连续人脸彩色图像序列进行处理的系统。该方法利用眨眼时开闭眼状态的灰度变化精确定位眼睛并提取眼睛模板,较之传统的人眼定位方法更简单、准确、稳定。在短时间内头部基本不动的情况下,计算出帧差累积图变化最大行与列以及双眼质心的位置,进而确定眼睛的位置并标准化眼睛模板。实验结果表明该方法能实时、准确、快速提取出人眼模板。

Eye location and Template Extraction Based on Accumulated Frame Differences

This paper presents an eye location method based on accumulated frame differences.A system,which can process real-time captured face video sequence with blinking status,is set up by using VFW in VC .Applying the grayscale difference between open and closed eye during blinking,the method can precisely locate the eye position and extract eye template.Compared with traditional eye location methods,it is simpler,more accurate and stable.Without moving head in a short time,vertical and horizontal position of the maximal grayscale and the mass centre of a pair of eyes on the accumulated frame differences image can be found.After finding these two points,one can locate eye and normalize the eye template.Experimental results demonstrate that the algorithm can accurately extract eye templates in real-time with short period of time.