Flexible and improved method for automatic Semantic Content extraction in videos

Now days, mostly video based applications are used. In those applications, we need to do semantic search and extraction of multimedia contents. To summarize video, video event detection is necessary. Using raw data and low level features user cannot fulfill his need, so deeper understanding of video is necessary. Here, Video Semantic Content Meta Model is presented, using which automatically extraction of objects, events and concepts is possible. This video semantic content model is based on Domain ontology and it uses spatial/temporal relationship in events and concepts. Extraction process starts with object extraction and then objects are used as input for event extraction. For object extraction, Depth Image Based Rendering (DIBR) is used. In DIBR, views for objects are considered so that object detection will be more correct. In proposed framework, we considered viewing angle of camera to improve extraction capabilities.