期刊名称：International Journal of Computer Science and Information Technologies
电子版ISSN：0975-9646
出版年度：2011
卷号：2
期号：2
页码：891-898
出版社：TechScience Publications
摘要：Despite advances in the archiving of digital video, we are still unable to efficiently locate and retrieve the portions that interest us. Video indexing by segmentation has been a proposed solution and several research efforts are seen in the literature. Segmentation alone cannot solve the problem of content based access to video. Recognition of the text in video has been proposed as an additional feature. Various methods were proposed for isolation of the text data from the documented video image among wavelet transforms have been widely used as effective tool in the text segmentation. This paper implements an efficient text algorithm for the extraction of text data from documented video clips. The implemented system carries out a performance analysis on various wavelets for the proper selection of wavelet transform with multilevel decomposition. Morphological operator is applied on the selected wavelet co- efficients for the text isolation and evaluates the contribution of decomposition levels and wavelet function to the segmentation results. This paper also implements neural networks for recognition of text characters from isolated text image and make tit editable.
