Non Parametric Statistical Analysis of Scene Activity for Motion-Based Video Indexing and Retrieval

This report describes an original approach for content-based video indexing and retrieval. We provide a global interpretation of the dynamic content of video shots without any prior motion segmentation and without any use of dense optic flow fields. To this end, we exploit the spatio-temporal distribution within a shot of appropriate local motion-related measurements issued from the spatio-temporal derivatives of the intensity function. These distributions are then represented by causal Gibbs models. The considered statistical modeling framework makes possible the exact computation of the conditional likelihood function of a video shot to belong to a given motion or more generally activity class. This property allows us to develop a general statistical framework for video indexing and retrieval with query by example. We build a hierarchical structure of the processed video base according to motion content similarity. We consider a similarity measure inspired from Kullback-Leibler divergence. Then, retrieval with query by example is performed through this binary tree using the MAP criterion. We have obtained promising results on a set of various real image sequences.