My former Ph.D. student Otávio Penatti (currently a post-doc at RECOD), my colleague Prof. Ricardo Torres, his student Fernanda Silva, our French cooperator Valerie Gouet-Brunet and I had a paper accepted at Pattern Recognition: “Visual word spatial arrangement for image retrieval and classification”. The work is a development upon Otávio’s Ph.D. work.
Here’s the abstract:
We present Word Spatial Arrangement (WSA), an approach to represent the spatial arrangement of visual words under the bag-of-visual-words model. It lies in a simple idea which encodes the relative position of visual words by splitting the image space into quadrants using each detected point as origin. WSA generates compact feature vector, making it useful for both image retrieval and classification. Experiments in the retrieval scenario show the superiority of WSA in relation to Spatial Pyramids. Experiments in the classification scenario show a reasonable compromise between those methods, with Spatial Pyramids generating larger feature vectors, while WSA provides adequate performance with much more compact features.
The preliminary version of this work, “Encoding Spatial Arrangement of Visual Words”, won the best paper award at the 16th Iberoamerican Congress conference on Progress in Pattern Recognition — CIARP 2011. Both preprints can be found at my publications page.