Study Of Video Caption Extraction Algorithm Based On Spatial-Temporal Information

Caption text routinely provides rich semantic information. Compared with other video features, information in caption text is highly compact and structured, thus is more suitable for efficient video indexing, therefore video caption based methods have attracted particular attention.This paper deals with how to make full use of spatial-temporal information to fulfill caption extraction with great efficiency and speed. In order to further video analysis, an algorithm of abrupt shot boundary detection based on fuzzy clustering neural network ( FCNN ) is proposed, and it has the advantages of high precision as well as robust to fast move. Caption segmentation is the key to the whole process, FCNN can also be utilized to locate caption region, however, the technique is time-consuming. Thus an improved projection segmentation method is presented, and the experimental results show that it is simple and practical, and fits for real-time processing. Because of the simplicity of background in an individual character figure, an approach of caption binarization based on an individual character is put forward. Eventually, after character segmentation, binarization and remaining background pixels elimination, a clear and legible caption is obtained, the fact is demonstrated by the result of character recognition.