Deep Palmprint Image Quality Assessment Network

Palmprint recognition has become a popular biometric topic in recent years of its several merits such as high security, easy collection, and non-invasive. However, most existing palmprint recognition methods usually performfeature extraction on palmprint images without assessing the quality of the palmprint images, which significantly affects the final recognition performance. Moreover, to the best of our knowledge, there is no image quality assessment (IQA) approaches for palmprint images quality assessment. In this paper, we put forward a deep palmprint image quality assessment network (DPIQAN) for the quality evaluation of palmprint images. Firstly, we proposed a ResNet-based network with pre-trained parameters to extract the quality features of palmprint images. Then, we engineer a regression layer to evaluate the quality of the palmprint images. We conduct extensive experiments on a widely used palmprint image database, showing that our proposed method outperforms previous state-of-the-art methods in palmprint IQA.