Enhanced shot boundary refinement for post-shot boundary detection

Automatic extraction of video information has become an essential step in the video indexing system. In order to extract information from the special effects incorporated in the sequences, multiple stages are integrated together for effective processing. Shot boundary refinement is the first stage of such a technique. Its purpose is to the exact the transition start and end. The information is used later to determine the types of special effect as well as extracting information within these effects. The algorithm detects the exact shot boundaries by first constructing a reference indicator and picks up the sharp gradient changes within such an indicator. This paper describes the background information and the theory behind such a technique.