Automatic Video Annotation using Bayesian Inference

Annotating videos manually is very costly and time consuming. Human being's subjective and different understanding often lead to incomplete and inconsistent annotations and poor system performance. So it is an important topic to automatically annotate a video shot. In this paper, we propose a new approach of automatically extracting a non-fixed number of semantic concepts for a video shot. The first step is to propose a simple but efficient method to obtain the semantic candidate set (SCS) based on visual features. The second step is to select the final annotation set from the SCS by Bayesian inference. Experimental results show that our method significantly outperforms NB algorithm and KNN algorithm in automatically annotating a new video shot, and is more robust than the two algorithms