Palmprint recognition based on weighted fusion of DMWT and LBP
Yunfeng Li, Yali Zhang · 2011
To obtain affluent features of the palmprint image, the weighted fusion method of Discrete Multiwavelet Transform (DMWT) features and Local Binary Pattern (LBP) features is proposed. This method fuses the global image features by DMWT and the local image features by LBP, which can overcome the limitation of the single feature extraction method, and synthesize two kinds of the image features. Principal Component Analysis (PCA) is used to solve the dimensional increase problem of this fusion for its powerful decrease ability. The implementations of this method are as follows: firstly, the LBP and DMWT are used to extract the features respectively; secondly, different weighted coefficients are multiplied to these two features; thirdly, the PCA is used to decline the dimension of the fused feature vector; finally, Euclidean distance is calculated to achieve the pattern recognition. Through this method, the best weighted coefficient can be found, and it will be used as the final weighted coefficient. The experimental results demonstrate the effectiveness of this palmprint recognition system.