Moving Object Detection for Content Based Video Retrieval

One of the challenges faced in video retrieval is accurately detecting and extracting the objects in motion. The properties of the moving objects from the query clip are obtained, represented quantitatively, and compared to the corresponding features of the moving objects in the videos stored in the database. If a large number of features match, it generally indicates that the objects are similar. This increases the probability that the database video is related to the query clip. This paper proposes a technique to observe the moving objects in the key frames of a video. In order to extract and determine a clear outline of these objects, an absolute difference technique is carried out. This is followed by several methods of filtering in order to smoothen the image and eliminate noise. The proposed method has been performed on several videos.