Rubbing Image Retrieval Using Deep Convolutional Neural Network

With the widespread use of digital imaging data, the size of image collections about Bronze Inscriptions rubbing is increasing rapidly. It becomes difficult to manage and query a specific image from these large databases, which motivates the need for image retrieval. In this paper, we proposed an effective content-based rubbing image retrieval (CBRIR) framework based on deep convolutional neural network (DCNN) for Bronze Inscriptions rubbing. Specifically, we extract discriminative local features for image retrieval using the activations of convolutional neural networks. We use cosine metric, Euclidean metric, and Hamming metric to measure similarity in the CBRIR framework. Experimental results show that our framework has an excellent accuracy of rubbing retrieval.