Video clustering based on the collaboration of multimedia clusterers

This paper addresses the task of video clustering, representing the videos as complex multimedia objects, which are an aggregation of heterogeneous data (text, images and audio) as a single unit. In support to the development of this task, this paper proposes a new method based on the collaboration of clusterers evaluating the multimedia resources present in videos. The new method, called Multimedia Collaborative Multi-clustering, is evaluated using the video set of MediaEval 2011, implementing a text clusterer with the speech transcription provided by the video set, an image clusterer with keyframes also provided by the video set, and an audio clusterer with the acoustic signal of the videos. A comparison against several solutions found in the literature demonstrates the feasibility of the proposed method, which creates clustering structures closer to the actual classification of the videos than the clusters produced by other solutions.