As an important biological feature, Palmprint recognition has been paid wide attention and obtained a large number of research achievements in the past few years, due to its security, convenience and stability. However compared with the traditional contact acquisition, the method with non-contact acquisition not only avoids the public environmental health problems, but also can effectively prevent the loss of the palmprint template and meanwhile, increase the security of the system. Nevertheless, there are some inherent defects in non-contact system, e.g. the palmprint image easily produces blurriness when it is captured by non-contact device, which possibly degrades the performance of system recognition. In order to solve aforementioned problem effectively, a blurred palmprint recognition based on Structure Feature is proposed. Firstly, we get a conclusion that structure feature of image is stable in the process of the image blurring by analyzing the blur theory. Next, the conclusion is proved from the point of different image layers by using decomposition model, and extracting the Structure Feature from the structure layer of palmprint image. Then, we used a non-overlapping sampling method based on Structure Ratio(SR) to further improve the discrimination and effectiveness of feature. Finally, Structural Similarity Index Measurement(SSIM) is used to measure the similarity of palmprint and judge the palmprint category for classification. Extensive experiments on the PolyU and blurred-PolyU palmprint databases are carried out, the recognition results of the proposed method are stable on the different palmprint databases and importantly, Equal Error Rate(EER: 0.8340%) of the proposed method is lower than other classical algorithms on blurred-PolyU palmprint database, which validate the effectiveness and superiority of the proposed method. |