Unsupervised Clustering of Cricket Videos using One Threshold Sequential Algorithm

In this paper we propose an approach for clustering a given cricket video by an unsupervised technique. The algorithm is based on one threshold basic sequential algorithm. We propose an unsupervised clustering technique to group the similar shots in a given cricket video into clusters such that the maximum mutual information is preserved. In this we use mutual information to measure the similarity between frames. The number of clusters formed depends on the shots present in the given video and varies for different videos. The algorithm has been tested for several cricket videos, From analysis, the proposed method was determined to be a efficient for video summarization clustering.