Text Semantics Based Automatic Summarization for Chinese Videos

This paper is concerned with the task of automatically summarizing videos, which is important for many video-related applications. Examples include video retrieval, accessing and categorization, et al. Our approach summarizes video semantically from the texts embedded in videos. To improve the accuracy, ant colony algorithm, maximum matching method, Singular value decomposition (SVD) are employed for locating text region, segmenting words and clustering sentences, respectively. Based on these methods, a prototype is developed. In an experimental evaluation on the real world video datasets, we show that our proposed approach could provide accurate, satisfactory performance.