Résumé :
This chapter explores specifically the clues that can assist human annotators in the temporal segmentation of circus performance video into acts. An online video retrieval application has been developed and published on the Circus Oz Living Archive website. The Circus Oz performance videos frequently are recorded in a single shot; lengthy, affected by camera operation and by fluctuating light intensity; captured in low-light environments; have variable audio quality and originate from multiple video formats. Black frames detection is a technique for finding the start time and the end time of a sequence of black frames occurring in a video. Image content analysis is another way to find that an act has ended in a circus performance video. The colour histogram comparison approach is used for analysing image similarity in the Circus Oz videos. The process for comparing two audio clips can generally be divided into two steps: audio feature extraction and audio distance calculation.

Localisation : Bibliothèque