This article provides a brief description on existing technique for Video segmentation on both compressed and uncompressed videos, this article also discuss about the algorithms for shot boundaries detection, and topics related to camera operation recognition. The resent researches in multimedia compression technology have led to the growth of internet, and use of digital video. Application such as digital libraries, video-on-demand, interactive TV, uses a large collection of video data. The main goal to sub divide video stream into a set of meaningful shots (segments) which can be used as basic element in indexing of video stream, where each shorts is represented by selecting key frames and indexed the retrieval is based on the feature vector of the query and stored video. A shot can be defined as a unbroken sequence of frames, There are mainly two types of shot transitions abrupt and gradual. Although there are many technique for combining two shots, most often fades and dissolves are been used. Camera operationrecognitionis one of the other basic important issues. As camera operationsusually explicitly reflect on how the attention of the viewer should be directed, the clues obtained are useful for key frame selection.
Temporal video segmentation: Most of the algorithms process uncompressed videos, similar measure can be done in case of images such as on comparing two images if they are dissimilar then there will be a cut, based on the metrics used difference between two frames the algorithms can be divided into: Pixel, block and histogram comparison. This research has also confirms the need for benchmark video sequences and unified evaluation criteria that will allow comparison and precise evaluation of the various techniques. The benchmark sequences should contain enough data for all the possible types of camera operations and shot transitions, including complex gradual transition
Download Video Segmentation Seminar Report and video segmentation ppt.