Video Retrieval Based On Weight Color Component And Particle Swarm Algorithm

With the vigorous development of multimedia technology and network technology in recent years,video data has been got the massively increased. And because of the complexity of video data,the traditional data processing methods can not meet demand.How to retrieve the video you needed from the massive video become very important,so content-based video retrieval has become the focus of multimedia research field.Content-based video retrieval analyzes the structure of the video,and divides the video into key frames,shot,scene and ultimately to the shot unit.According to the video example submitted by the user,finding the video clips with similarity in the video database,and the search results are given according to the level of similarity.This paper comprehensively analyzes the previous content-based video retrieval method, and proposes video segmentation based on weight color component and key frame extraction based on particle swarm algorithm,the details are as follows:(1) Shot boundary detection.This paper proposes the shot detection algorithm based on the weight color component after studying the existing shot boundary detection method.This method chooses HSV color space as the feature description space,including three subspaces,they are color H subspace,saturation S subspace and brightness V subspace.Using the three subspaces reflect the different feature in the shot changing,adaptively distributes the subspace weight,highlighting the subspace which changes highly.This can make the shot boundary and the shot internal movement easier to differentiate.Finally,the experiments verify the effectiveness of the algorithm.(2) The key frame extraction. Considering that the main reason of key frame extraction is the video movement.This paper divides the movement into two kinds: local motion and global motion.First,extracting the optical flow features under the Model Note from the upper left corner area as the image of global motion features,at the same time,extracting the color features from the middle area as the image of local motion features.Then the global motion and the local motion are imerged as the image's feature vector.The paper analyzes a lot of video data,and chooses five typical video's feature vector.We use feature vector of typical video as elementary particles to adaptively extract key frame by particle swarm algorithm. Experimental results show that the key frame extraction algorithm can effectively reflect the major events of the shot and the redundancy is in good condition.(3) The design and implementation of the prototype system. In order to verify the validity of the above methods.we use object-oriented design methods and Visual C++6.0 and OpenCV1.0 for the development of tools, implementing content-based video retrieval prototype system. The prototype system includes:video pre-processing, boundary detection, key frame extraction, feature extraction, shot similarity disputation and video retrieval.