Research on Product Image Retrieval with Local Features

A study is made on the effects of different local features on product image retrieval. First, Corners are detected using Harris-Laplace corner detector and get regions at corners' scale. Then, extracting local features at each region, such as SIFT. Finally, matching feature points and retrieval image using match result. Meanwhile, A Comparison is made on the invariants of different local features on various affine transformation, and the effects on product image retrieval. The experiment and analysis are some reference values to the local feature selection on product image retrieval.