A Document Image Retrieval Method Based on Multi-feature Fusion

With the rising popularity of mobile devices, recent years have witnessed a growing interest in document image retrieval (DIR). In conventional Bag-of-Visual-Words (BoVW) based document image retrieval method, only SIFT or SURF feature is used to locate feature points and produce a codebook, which has a low discriminative power. Though several multiple feature based BoVW methods are proposed, these methods are based on multiple codebooks. However, some features can not train a codebook, such as connected-component feature. So this paper propose a new multi-feature fusion method, considering these features. Experimental results show that the proposed method has better performance.