Performance analysis of multiple classifier fusion for semantic video content indexing and retrieval

In this paper we compare a number of classifier fusion approaches within a complete and efficient framework for video shot indexing and retrieval1. The aim of the fusion stage of our sytem is to detect the semantic content of video shots based on classifiers output obtained from low level features. An overview of current research in classifier fusion is provided along with a comparative study of four combination methods. A novel training technique called Weighted Ten Folding based on Ten Folding principle is proposed for combining classifier. The experimental results conducted in the framework of the TrecVid'05 features extraction task report the efficiency of different combination methods and show the improvement provided by our proposed scheme.