Effects of Different Types of Noise in Palmprint Image Classification Using SURF Descriptors

In this paper we present some results obtained in experiments made on the effect of different types of noise in palmprint image classification using SURF descriptors. Even the use of SURF features in noise image classification led to good results, it is usually a vast time-consuming procedure, depending on parameter values used in SURF keypoints generation. To obtain a shorter computing time in a noise image classification we used LBP texture features to select a subset of candidate images and then process the subset using SURF keypoints. This proposed approach has led to very similar results in a shorter computing time than direct SURF approach. Experiments were made on four well-known palmprint databases and the noise images were not preprocessed.