摘要
根据皮肤纹理的深浅、节点数及纹理粗细等主要特征,采用小波包及二维傅里叶变换分析方法和数学形态学分析法实现了皮肤的特征提取。利用小波包分解得到的空间频率信息对二值化纹理图像进行空域滤波去噪,利用基本竞争型人工神经网络进行纹理分类。结果表明,这一方法可获得较理想的分类结果。
The wavelet packet transform, two-dimension Fourier transform and mathematical morphology are employed to realize extraction of features of skin texture according to main texture features such as skin surface depth?count of nodes and thickness of skin texture. The space frequency information obtained from wavelet packet transform and two-dimension Fourier is utilized to achieve space noise-reducing, and a counter propagation network is used to classify sample skin textures. The results of experiments show that the cataloging outcome of the method is quite satisfying.
出处
《光学技术》
CAS
CSCD
2004年第4期467-469,472,共4页
Optical Technique
关键词
皮肤纹理
特征提取
小波
数学形态学
skin texture
feature extraction
wavelet
mathematical morphology
