29-09-2016, 03:45 PM
1456873031-Aliteraturesurveyonvideosummarization.docx (Size: 103.66 KB / Downloads: 8)
Abstract-Over the past couple of years,there has been agreat increase in amount of video based content are created .There have been tremendous needs of video processing application to deal with the massive collection of video content data .This research paper covers the study of different techniques of video classification which is based on key frame.
Introduction- massive growth in video content cause problem of overload and management of data to manage that video content data we have need of video classification .video classification is process can deal with massive video base content and can represent the summarization of video (video classification) data. Video summarization aims succinct representation to enable a quick browsing of collection of massive video database and one of the research areas of interest is video summarization or classification is process of creating and presenting a meaningful abstract view of entire video with in short period of time two type of video classification are availablein the literature i] key framed based and ii]video skimming . This paper also preview some of the recent work on content based multimedia information retrieval and video classification representation.
Video Anatomy –
A video is synchronous sequence of collection of frames .video classification has basic unit which is a frame video is collection of different scene and scene is collection different shots and shots are nothing but collection frames . For the summarization of the video can be done by the two different way which are as mentioned as i] Video classification by key frame ii]video classification by video skimming.
I]Video classification by key frame-
Video classification by key frame can also called representative frame still image abstracts ,R-frame and a collection of salient image extracted from the video source at the implementation of key frame based algorithm have some challenges like.
– Redundancy frame with minor difference are selected as key frames.
– To create clustering at the multiple changes in video.
II] Video classification by video skimming-
Video classification by video skimming techniques also called symmetry sequence, moving image abstract ,moving story board . Original video is divided into various part, which is a video clip with short time period .The movie trailer is the good example of video skimming.
A] General ClassificationMethodology-
Video classification is generally based on key frames. Important step of key frame extraction algorithm by using low-level visual features
B] Allocation of video into shots –
First step in the classification of video is to allocate the video into numbers of shots. Large change in the video frame content cause at the shot boundaries for the representative of the frame content the histogram of frame is used entropy of histogram can be used instead of a histogram.
C]Computing classification parameters for each frame –
Mostly motion audio level,colour moments are computed for each frame for the classification
(a) Motion-
This is most important criteria used for classification of a video shots carry more important information about the video that express what is happening in the video.
Motion in the frame can be compute by server approaches .one of the most popular approach is to compute the optical flow by using horn and schunk’s optical flow algorithm.
(b) key-frame Extraction-
This is the anotherunit to decide which shot is most important for the collection of key-frame that shots contains.
Key frame is he frame that represent the video content specific pattern . colour difference between key frame and it is succeeding frames is not large until the next key frame arrives and colour difference between them in quite large then key frame and the all others neighbour frame are similar to it.
For the key frame extraction we calculate the first three moments from the moment of the previous frame.
Video classification Presentation-
Video classification represent by two method static and dynamic presentation, And these presentation can present condensed form of the similar key frame those are contained together. Video classification are commonly present as asset of static key-frame or dynamic video skim.
(a) static presentation –
This the most common method for video classification presentation technique. This technique also called storyboard, this is nothing but the static grid of extracted key frame according to recent study on evolution of video classification technique (west-man 2010) the story-board has complete strength to give video content.
(b) Dynamic presentation –
The anotherpresentation technique that transform the original video into short version. Skimming of the video are nothing but short video cut-out from the original video. Dynamic video skimming also support the recognition of specific objective pattern in the content.
Future work–
This Techniques can be in that system which can use to recognition of various features such as Face Detection. The location of the place can be traced from where the alert has been obtained. SMS or Alarm system can be employed if a particular feature is obtained. This will effectively reduce the crimes and provide an immediate response system at emergency. Together all this technology features forms a basis of improved security systems. In future, efficient video Processing tools will change the face of this security system.
Conclusion –
Video summarization plays important role in many video applications. This survey on video classification and video classification techniques are carried out and also shows presentation of classification which is useful for the data presentation in in specific manner which is useful for quick data browsing from database