基于光谱融合的掌纹精细纹路提取

Palmprint fine line feature extraction based on spectral fusion

针对掌纹中浅而细小的精细纹路,提出一种光谱融合的非下采样Contourlet变换(NSCT)与数学形态学细化相结合的掌纹线特征提取方法。该方法先对不同光谱下的掌纹图像分别进行NSCT分解,根据精细纹路特点处理子带系数矩阵,结合光谱特点融合各系数,逆变换后进行形态学细化得到掌纹精细纹路特征。使用自行采集的多光谱掌纹图像实验表明,本文方法提取出的浅而细小的掌纹精细纹路更清晰丰富。

Classical palmprint feature extraction algorithms can not extract the shallow fine lines in a better way. To solve this problem an algorithm based on Non Subsampled Contourlet Transform （NSCT） and mathematical morphology multispectral palmprint line feature extraction is proposed. First, NSCT is used to obtain multi-directions, multi-scales and sub-bands are selected and strengthened according to the shallow fine line characteristics. Then, in light of the spectral characteristics, the coefficients are fused and NSCT inverse transformation is made. Finally, using morphological thinning treatment, the shallow fine palm lines are obtained. Experiments using multispectral palmprint image acquired from self-collected palmprint database show that the propose algorithm can extract shallow fine lines more clearly and abundant.