A classification algorithm for hologram label based on improved sift features

Abstract

Hologram label can present different images under different light condition. Thus, it is difficult to recognize a hologram label with traditional methods. In this paper, we propose a classification algorithm for hologram label based on improved SIFT features. Firstly, a multi-illumination sample space is constructed by collecting images from one hologram label under different illumination condition. Secondly, the SIFT features are extracted from different samples in the multi-illumination sample space. Thirdly, some stable feature points are obtained after matching and clustering steps. Finally, the class of a testing hologram label is determined by the number of the matched SIFT points. Experimental results show that our method has good accuracy and recall rate, especially the ambiguous images can also be recognized.