Abstract
The emergence of multimedia information, such as video, has presented a problem to traditional information management systems which were not designed to handle media other than standard alphanumeric data. To handle a continuous media such as video, one should segment it into basic units to be used for identification and indexing. The main problem in most existing techniques for locating the boundaries between the basic units or scenes is the selection of thresholds with which the interframe differences are compared. Various changes might happen within a single scene, which include changes occurring as a result of camera operations, object movements, flashes, etc. On the other hand, various similarities between different scenes might occur. Therefore, selection of any threshold value will result in low detection rates due to detecting false cuts or missing true cuts. In this paper we propose a detection mechanism that understands and detects various changes within a scene in order to detect the cuts between different scenes. These cuts may represent sharp or gradual cuts. This mechanism depends on the hue value flow generated as a result of camera operations, object movements or cuts.