摘要
采用改进的BP算法,实现人脸图像的边缘检测。构造了学习样本,并在较短时间内训练得到权值矩阵,从而实现二值图像边缘检测。在此基础上,综合灰度图像各位面的边缘提取结果,实现对灰度图像的边缘检测。利用ORL人脸数据库图像的实验,证明了将得到的权值矩阵用于人脸图像的边缘检测,泛化性较好,抗噪能力强,能得到较为连续精细的边缘。
An improved BP algorithm is proposed for edge detection of face images. Firstly, the samples are constructed and the weighted matrix is got by exemplifying in a short time span. Then the edge of binary images can be detected through the process. Furthermore, first divide each gray-scale image into 8 binary planes with different gray level, and then naturally the images surface up through synthe- sizing the edge of each binary plane. Applying the method to face images' edge detection, fine continuous edge is acquired, the virtues of which lies in its strong resistance against noise pollution and is widely adopted in practice.
出处
《计算机工程与设计》
CSCD
北大核心
2007年第21期5181-5182,5260,共3页
Computer Engineering and Design