基于梯度模板匹配的眉毛识别方法

摘要

针对基于快速傅里叶变换的眉毛识别方法(FFTER)识别速度较慢的问题,设计了一种新的基于梯度模板匹配的眉毛识别方法(GTMER) .该方法识别速度更快,更能够适合于实时应用.首先,利用图像梯度特征,快速匹配出纯眉毛模板的最相似区域;然后,通过小范围内的精准调整,得到一个准确的匹配位置;最后,利用傅里叶频谱距离,将待处理眉毛图像识别出来.实验表明:GTMER的识别速度最高比FFTER提升约38%,同时在BJUTED眉毛数据库上的识别正确率也达到98.12%.%Aimed at slow running speed for fast Fourier transform eyebrow recognition method (FFTER),we pres-ent gradient template matching eyebrow recognition method (GTMER),a real-time method in matching-recog-nizing framework. It firstly uses image gradients to match the most similar sub regions of templates very fast,and then obtains the adjustment matching position in a small area of the most similar sub region,finally recognizes the input eyebrow image with a Fourier spectrum distance. Experimental results show that GTMER runs about 38% fast-er than FFTER and achieves the accuracy of 98.12% on the BJUTED eyebrow database.