Improving Image Similarity With Vectors of Locally Aggregated Tensors

David Picard (1) , Philippe-Henri Gosselin (2)

1 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes
2 MIDI - Multimedia Indexation and Data Integration

Within the Content Based Image Retrieval (CBIR) framework, three main points can be highlighted: visual descriptors extraction, image signatures and their associated similarity measures, and machine learning based relevance functions. While the first and the last points have vastly improved in recent years, this paper addresses the second point. We propose a novel approach to compute vector representations extending state of the art methods in the field. Furthermore, our method can be viewed as a linearization of efficient well known kernel methods. The evaluation shows that our representation significantly improve state of the art results on the difficult VOC2007 database by a fair margin.

Keywords: Image classification; Image representation; Bag of Words

Domains: Machine Learning [cs.LG]

David Picard, Philippe-Henri Gosselin. Improving Image Similarity With Vectors of Locally Aggregated Tensors. 2011 IEEE International Conference on Image Processing (IEEE ICIP2011), Sep 2011, Brussels, Belgium. pp.669 - 672, ⟨10.1109/ICIP.2011.6116641⟩. ⟨hal-00591993⟩