In this talk, I will first show that the search problem can be cast into a source coding framework, where the database vectors are approximated by quantization. The Euclidean distance between a query vector and a database vector is estimated in an asymmetric manner based on the quantized database descriptors. The method is advantageously combined with an inverted file to avoid exhaustive search, and used either for local or global descriptors.
I will then discuss some recent improvement of this method in a context of precise image matching, and show the importance of the nearest neighbor search accuracy for image search.