Video, a much richer media compared with text, is now widely used in many applications like broadcasting, education, publishing and military intelligence. However, the effective use of video content is beyond our grasp. Retrieving specific information from a very large digital multimedia database or performing effective video content navigation is still very difficult. We have been dealing with textual documents for thousands of years and found comprehensive ways to index, search and incorporate its contents. For video contents, many research efforts have been devoted to content-based video indexing and retrieving. The result till now is still far beyond our satisfaction....[ Read more ]

Video, a much richer media compared with text, is now widely used in many applications like broadcasting, education, publishing and military intelligence. However, the effective use of video content is beyond our grasp. Retrieving specific information from a very large digital multimedia database or performing effective video content navigation is still very difficult. We have been dealing with textual documents for thousands of years and found comprehensive ways to index, search and incorporate its contents. For video contents, many research efforts have been devoted to content-based video indexing and retrieving. The result till now is still far beyond our satisfaction.

In this work, we propose an enhanced system model to extract story units from MPEG movies. The model is an extension of the work of Minerva, Boon-Lock and Bede [9]. Based on the low level features, including video shots, key frames, colours, blobs and scene transition graphs, we have attempt to extract high-level semantic structure of a movie. The architecture of the system, techniques to perform camera break detection, identify key frames, match frames with colour and blob information are discussed. The effectiveness of the approach and required knowledge are presented.