XML representation of digital videos for visual data mining applications

A novel approach utilizing the eXtended Markup Language (XML) for representing digital videos is proposed. This new method is based on segmenting the video sequences into real-world objects and then translating the objects into a hierarchy that is easily represented by XML. The algorithm segments each frame of the video sequence into consistent objects without using MPEG by Leonardo Chiariglione (1999) motion vectors or intra-block information, thus eliminating common errors contributed from these computed sources. This efficient identification and representation of corresponding objects between frames allows for a host of visual data mining applications utilizing digital libraries, databases and Internet image searches. This new algorithm is tested on various video segments (pans, zooms, dissolves, close-ups, and multiple-object motion) with results included.