Research And Implementation Of Key Frame Extraction Algorithm Based On Information Entropy

With the rapid development of Internet technology and multimedia information,multimedia data,especially video data,has become one of the most important ways for people to obtain information in their daily lives.However,it has become a scientific problem to efficiently and quickly extract information required by users from massive video data.Many years of research in the field of video retrieval show that the scientific and rational use of keyframe extraction technology can reduce the amount of data generated in the video retrieval process,save video processing time,and improve work efficiency.Shot boundary detection is the precondition and basic step of key frame extraction.Des igning an effective and reasonable lens boundary detection method has an important influence on key frame extraction.On the basis of the predecessors,this paper proposes an improved lens boundary detection algorithm: first through the lens pretreatment,and then calculate the dynamic difference of mutual information entropy,finally,the variance of the adjacent frame luminance histogram can be used to effectively detect the abrupt and gradual lens frame.Experiments show that the improved lens boundary detection algorithm in this paper reduces the amount of computation in the lens boundary detection process,thereby reducing the time overhead of the algorithm.In this artic le,key frame extraction is the focus of research.By studying existing keyframe extraction algorithms,we have found that the speed and time for extracting key frames tends to increase significantly with the increase in the amount of video data.This does not satisfy the current demand for video retrieval.Based on the shortcomings of current key frame extraction algorithms,this paper proposes a key frame extraction algorithm based on mutual information entropy and SUSAN measure.The algorithm first uses the shot boundary detection method to detect the shot boundary in the video,and divides the video based on the change of the shot into video clips.Then,the candidate key frames in the video clip are extracted using the dual feature amounts of mutual information entropy and SUSAN operator.Finally,the edge matching algorithm is used to eliminate the redundant frames in the candidate key frame,so as to extract the key frame set representing the main content of the video.In order to reduce the time overhead of the key frame extraction process,this method uses the CUDA parallel model architecture to utilize the parallel mechanism in feature extraction,edge detection,and redundancy elimination.Finally,the effectiveness of the key frame extraction algorithm in this paper is verified by experiments.