基于KL变换的指纹特征提取应用研究

Research on the Application of Fingerprint Feature Extraction Based on KL Transform

KL变换是一种适用于任意概率密度函数的正交变换,它能消除各分量之间的相关性,根据协方差矩阵特征值和特征向量有效地进行信息压缩等。相同类的指纹图像的特征码具有较强的相似性,不同类指纹图像的特征码却有一定差异。采用对特征码进行KL变换降维,得到的新的特征码仍然具有同样的相似性和差异性。证明可以通过KL变换后的特征向量进行指纹识别是可行且具有一定意义和研究应用价值。

KL transformation is a probability density function for arbitrary orthogonal transformation. It can eliminate the correlation between the various components, effectively compressing and other information based on the covariance matrix of eigenvalues and eigenvectors. The fingerprint image of the same class signatures with strong similarity, different types of fingerprint pattern there is a difference image. KL transform on the signature dimensionality reduction, the new signature is still with the same similarities and differences. Fingerprint proof can be transformed by eigenvectors KL is feasible and has a certain significance and application value.