摘要： 介绍了基于Gabor小波的眼睛和嘴巴检测问题。首先利用肤色概率模型、人脸图像的梯度信息以及运动模型提取出人脸器官的可能特征点,接着使用一系列规则筛选出最可能是眼睛对的特征点对,然后采用Gabor小波和支持向量机(SVM)来对可能的眼睛对进行有效的验证,最后通过眼睛的位置来找到嘴巴的精确位置。实验结果表明,该系统具有较好的眼睛和嘴巴检测效果,还可以通过再学习进一步提高系统的检测性能。

关键词: Gabor小波, 眼睛对验证, 嘴巴检测, 支持向量机

Abstract: A Gabor wavelet-based eyes and mouth detection system is presented.To extract the potential feature points in face organs,this approach combines the skin color model,grayscale gradient clue,and motion model.And then the position-pair of highest likelihood,which is determined from a set of rules,is taken into account;all of the probable eyes-pairs by Gabor wavelet and SVM can be verified.Finally,the position of mouth through the position of eyes can be detected.The result of experiment shows that the proposed system has higher rate of right eyes and mouth detection.In addition,the performance of system can be improved by further learning.

中图分类号: 

• TP391.41