Video copy detection: a comparative study

摘要: This paper presents a comparative study of methods for video copy detection. Different state-of-the-art techniques, using various kinds descriptors and voting functions, are described: global descriptors, based on spatial temporal features; local spatial, as well spatio-temporal information. Robust functions is adapted to these techniques enhance their performance compare them. Then, dedicated framework evaluating systems proposed. All the tested compared within same framework, by robustness under single mixed image transformations, different lengths segments. We discuss each approach according transformations applications considered. Local demonstrate superior over ones, when detecting copies subjected transformations.