Key Techniques Of Content-based Video Analysis

With the rapid development of mobile network and information technology,the amount of multimedia video data grows explosively with complicated structures,various types and high definition,which brings serious challenges of how to efficiently storage,manage and analyze video data.In the context,content-based video analysis technology came into being and become a hot research highlight in video information analysis.The main work of this paper are improving algorithms of shot boundary detection and key frames extraction.On the basic analysis of video structure,a new method of shot boundary detection is proposed to overcome the defects of sensitive to camera movement and light change.This method makes use of the global feature of HSV color histogram information and utilize DPHA feature to compensate for the lack of spatial location information previously.Experimental results indicate the improvement of the precision and recall ratio of the proposed method.Then,in order to cover the shortage of video structure information in gradual shot boundary detection,a method based on shot transition model is proposed.The algorithm utilizes the characteristics of the invariable difference between the adjacent frames to detect gradual shot boundaries.This method is simple but effective which can eliminate the interference of large object motion in video frames.The experimental simulation shows the better identification of gradual shot boundaries and the improvement of the precision and recall ratio of the proposed method.Finally,based on the former study of video data segmentation above,an improved algorithm based on multi-feature cluster is put forward.This algorithm combines the feature of HSV color information and image edge information to be the decision-making cluster condition,and then carry out an adaptive cluster method which will overcome the limit of clustering initializing constraint and improve the flexibility of the proposed algorithm.Experimental results indicate that under the condition of the compression rate remaining at a high level,the key frame extracted by this algorithm can achieve better fidelity index than others,and the algorithm proposed is more reliable and efficient for different kinds of video...