基于四元数并行融合的手指纹理区域识别

Quaternion parallel fusion based finger texture identification

本文在四元数并行融合的基础上提出了一种基于四元数性质的并行融合手指纹理区域识别新算法,即提出了基于四元数模性质的QM匹配分数,并对手指纹理这一可靠生物特征的使用作了可行性分析。首先,从手指纹理区域ROI提取出4种二维小波分解系数;其次,按顺序将4种小波系数并行融合成四元数;最后,提出基于四元数模的QM匹配分数用于手指纹理识别。正常条件下,该算法效率上优于其他算法;噪声干扰条件下,该算法较其他算法具有更好的抗噪性。

Based on quaternion parallel fusion, this paper proposes a novel quaternion property based method for finger texture identification, which is quaternion modulus （QM） based matching score. Also, a discussion about the usability of the method as finger texture identifier is given. Firstly, 2D wavelet decomposition coefficients are captured from the finger texture. Secondly, a quaternion is constructed, which is a representation for 4-feature parallel fusion. Finally, a QM matching score is conducted for identification. The advantage of the algorithm is fast speed and anti-noise robustness.