Visual query expansion with or without geometry: Refining local descriptors by feature aggregation
作者：
Highlights：
• Novel query expansion method aligned with Hamming Embedding which individually matches features.
• Refinement of local description instead of global one in the previous query expansion methods.
• Descriptor aggregation providing more compact and more robust local representation.
• No need for spatial matching to identify good candidates for query expansion.
摘要
Highlights•Novel query expansion method aligned with Hamming Embedding which individually matches features.•Refinement of local description instead of global one in the previous query expansion methods.•Descriptor aggregation providing more compact and more robust local representation.•No need for spatial matching to identify good candidates for query expansion.
论文关键词：Image retrieval,Query expansion,Hamming embedding
论文评审过程：Received 3 October 2013, Revised 26 March 2014, Accepted 2 April 2014, Available online 12 April 2014.
论文官网地址：https://doi.org/10.1016/j.patcog.2014.04.007
