摘要
研究了一种基于曲波域的指纹图像预处理方法。首先将指纹图像在曲波域中分解,然后用Gabor滤波器来处理粗尺度系数,这些系数是原图像的近似值;同时,在细尺度系数上使用软阈值函数减少沿着脊线方向的噪声。再将重构以后的图像二值化,最后使用基于模板的脉冲耦合神经网络(PCNNs)细化算法细化二值图像,得到指纹的骨架图像。实验结果表明,该方法优于传统的基于Gabor滤波器的指纹图像预处理方法。
A method based on curvelet domain for pre-processing the fingerprint images is proposed.First,the fingerprint image should be decomposed in curvelet domain,than we use Gabor filters on coarse scale coefficients which values are the original image approximation;meanwhile,we use a soft threshold on fine scale coefficients to decrease the effect of noise along the ridge directions.Then,after binarizing the reconstructed fingerprint image,we use a image thinning method based on template-based pulse-coupled neural networks（PCNNs） to thin the binary image,and get the skeleton image.The simulation results show that the proposed fingerprint pre-process method is better than the traditional method which is based on Garbor filters.
出处
《激光与红外》
CAS
CSCD
北大核心
2010年第4期442-446,共5页
Laser & Infrared
基金
国家自然科学基金项目(No.60574051)资助
关键词
曲波域
GABOR滤波器
软阈值
模板
PCNNs
curvelet domain
Gabor filters
soft threshold
template
pulse-coupled neural networks（PCNNs）
