Deep convolutional features can provide rich semantic information for image content description. In order to highlight the object content in the image representation, the multi-center convolutional feature weighting method was proposed based on the relationship between high response positions and object regions. Firstly, the pre-trained deep network model was used to extract the deep convolutional features. Secondly, the activation map was obtained by summing the feature maps in all the channels and the positions with top few highest responses were considered as the centers of the object. Thirdly, the number of the centers was considered as the scale, and the descriptors corresponding to different positions were weighted based on the distances between these centers and the positions. Finally, the image representation for image retrieval was generated by merging the image features obtained based on different numbers of centers. Compared with Sum-pooled Convolutional (SPoC) algorithm and Cross-dimensional (CroW) algorithm, the proposed method can provide scale information and highlight the object content in the image representation, and achieves excellent retrieval results in the Holiday, Oxford and Paris image retrieval datasets.

The spatial pyramid matching method provides the spatial information by splitting an image into different cells. However, spatial pyramid matching can not match different parts of the objects well. A hierarchical image representation method based on Color Level (CL) was proposed. The class-specific discriminative colors of different levels were obtained from the viewpoint of feature fusion in CL algorithm, and then an image was iteratively split into different levels based on these discriminative colors. Finally, image representation was constructed by concatenating the histograms of different levels. To reduce the dimensionality of image representation, the Divisive Information-Theoretic feature Clustering (DITC) method was used to cluster the dictionary, and the generated compact dictionary was used for final image representation. Classification results on Soccer, Flower 17 and Flower 102 datasets, demonstrate that the proposed method can obtain satisfactory results in these datasets.