Abstract:

The coding methods of the traditional visual features adopted in current image retrieval approaches lack sufficient learning ability and have no strong feature expression ability.In addition, due to the high dimensionality of visual features, a large amount of memory is consumed, thus reducing the performance of image retrieval. In this paper, an image retrieval algorithm with end-to-end training based on deep and improved hashing method was proposed and designed. The proposed algorithm combined the high-level features extracted by CNN with Hash function and learned Hash codes with expression ability to perform large-scale image retrieval in low-dimensional Hamming space. The experimental results on two main datasets showed that the retrieval performance of the proposed method was superior to that of some state-of-the-art ones.

Key words: image retrieval, convolutional neural network, Hashing method, visual feature, Hamming space, feature coding, feature dimension