Image retrieval systems in commercial stock photography websites. Reverse image searches effectiveness

The aim of this study is to analyze the effectiveness of image retrieval in stock photography websites when a reverse search (an image query by a previous one) is performed. For this purpose, stock photography sites image retrieval systems are analysed; a set of simple images from the photographic genres (landscape, still life and portrait) is selected and a search of them in a sample of sites is processed. The results are compared with the reference images according to the attributes of texture, shape and predominant colours. Finally, a complementary textual search is also carried out for comparative purposes. It is concluded from the results and their analysis that reverse search is only offered by fee-based stock fotography sites, that it works better on images of some genres than others and that, in general, searches using sample images can be useful and effective for a first non-semantic query.