Object localization in metric spaces for video linking

Objects often constitute the desired level of access for browsing and retrieval in video databases. We present an approach to create links between video segments that contain objects of interest, based on video structuring, object definition, and stochastic localization in the video structure. Localization is formulated in the metric mixture model framework, which allows for the joint probabilistic modeling of a (user-defined) set of color appearance exemplars and their geometric transformations. Candidate configurations are drawn from a prior distribution using importance sampling, guiding the search towards regions of the configuration space likely to contain the true configuration, thus avoiding exhaustive processing, and evaluated using Bayes' rule. Experimental results on a small database of real colored objects extracted from home videos, with variations of scale and pose across video shots, show promising performance of the method.