Query-adaptive Image Retrieval by Deep Weighted Hashing

The hashing methods have attracted much attention for large scale image retrieval. Some deep hashing methods have achieved promising results by taking advantage of the better representation power of deep networks recently. However, existing deep hashing methods treat all hash bits equally. On one hand, a large number of images share the same distance to a query image because of the discrete Hamming distance, which cannot provide fine-grained retrieval since the ranking of these images is ambiguous. On the other hand, different hash bits actually contribute to the image retrieval differently, treating them equally greatly affects the image retrieval accuracy. To address the two problems, we propose the query-adaptive deep weighted hashing (QaDWH) approach, which can perform fine-grained image retrieval for different queries by weighted Hamming distance. First, a novel deep hashing network is designed to learn the hash codes and corresponding class-wise hash bit weights jointly, so that the learned weights can reflect the importance of different hash bits for different image class. Second, a query-adaptive image retrieval method is proposed, which rapidly generate query image’s hash bit weights by the fusion of the semantic probability of the query and the learned class-wise weights. Fine-grained image retrieval is then performed by the weighted Hamming distance, which can provide more accurate ranking than the original Hamming distance. Extensive experiments on 3 widely used datasets show that the proposed approach outperforms state-of-the-art hashing methods.