CAJ | 학술논문

关键帧可以有效减少视频索引的数据量，是分析和检索视频的关键。在提取关键帧过程中，为了解决传统聚类算法对初始参数敏感的问题，提出了一种改进的基于视频聚类的关键帧提取算法。首先，提取视频帧的特征，依据帧间相似度，对视频帧进行层次聚类，并得到初始聚类结果；接着使用K-means算法对初始聚类结果进行优化，最后提取聚类的中心作为视频的关键帧。实验结果表明该方法可以大幅提高关键帧的准确率和查全率，能较好地表达视频的主要内容。
관건정가이유효감소시빈색인적수거량，시분석화검색시빈적관건。재제취관건정과정중，위료해결전통취류산법대초시삼수민감적문제，제출료일충개진적기우시빈취류적관건정제취산법。수선，제취시빈정적특정，의거정간상사도，대시빈정진행층차취류，병득도초시취류결과；접착사용K-means산법대초시취류결과진행우화，최후제취취류적중심작위시빈적관건정。실험결과표명해방법가이대폭제고관건정적준학솔화사전솔，능교호지표체시빈적주요내용。
Key frame candramatically reduce the data of video indexing, and it is the fundamental processes in video analysis and video retrieval.In order to solve the problems that the traditional clustering algorithm is sensitive to the initial parameter in key frame extraction process, we propose an improved key frame extraction algorithm based on video clustering.Firstly, we extract the features of video frames. And the hierarchical clustering algorithm is used to obtain an initial clustering result, according to thesimilarity between two video frames.Then, K-means algorithm is conducted to optimize the initial clustering result and obtain the ifnal clustering result. Finally, the center frame of each clustering is extracted as key frame. Experimental results show that the precision and recall ratio of our proposed algorithm are greatly improved. The key frames extracted by our algorithm are better to express the primary content of video.