Efficient concept detection by fusing simple visual features
Duy-Dinh Le, Shin’ichi Satoh · 2009
Concept detection is one of the important tasks in video indexing due to its importance to bridging the semantic gap in multimedia retrieval. Many methods have been proposed for this task, however finding a method which can generalize well for a large number of concepts and is scalable for processing huge video databases is still challenging. In this paper, we introduce a general framework for efficient and scalable concept detection by fusing SVM classifiers trained by only simple visual features such as color moments, edge orientation histogram and local binary patterns. We evaluate the proposed framework for detecting a large number of concepts on various TRECVID datasets with hundreds of hours of video. Experimental results show that the proposed framework achieves good performance with a small computational cost.