Video Scene Detection by Multimodal Bag of Features

Abstract

Recent advances in technology have increased the availability of video data, creating a strong demand for efficient systems to manage this kind of content. To make efficient use of video information, first, the data have to be automatically segmented into smaller, manageable and understandable units, like scenes. This article presents a new multimodal video scene segmentation technique. The proposed approach combines Bag of Features based techniques (visual and aural) in order to explore the latent semantics obtained by them in a complementary way, improving the scene segmentation. The achieved results showed to be promising.