Fast and robust duplicate image detection on the web

Social media intelligence is interested in detecting the massive propagation of similar visual content. It can be seen, under certain conditions, as a problem of detecting near duplicate images in a stream of web data. However, in the context considered, it requires not only an efficient indexing and searching algorithm but also to be fast to compute the image description, since the total time of description and searching must be short enough to satisfy the constraint induced by the web stream flow rate. While most of methods of the state of the art focus on the efficiency at searching time, we propose a new descriptor satisfying the aforementioned requirements. We evaluate our method on two different datasets with the use of different sets of distractor images, leading to large-scale image collections (up to 100 million images). We compare our method to the state of the art and show it exhibits among the best detection performances but is much faster (one to two orders of magnitude).