期刊名称：International Journal of Computer Science and Network
印刷版ISSN：2277-5420
出版年度：2019
卷号：8
期号：2
页码：144-147
出版社：IJCSN publisher
摘要：Searching for a Video in World Wide Web has augmented expeditiously as there’s been an explosion of growth in video on
social media channels and networks in recent years. At present video search engines use the title, description, and thumbnail of the video
for identifying the right one. In this paper, a novel video searching methodology is proposed using the Video indexing method. Video
indexing is a technique of preparing an index, based on the content of video for the easy access of frames of interest. Videos are stored
along with an index which is created out of video indexing technique. The video searching methodology check the content of index
attached with each video to ensure that video is matching with the searching keyword and its relevance ensured, based on the word count
of searching keyword in video index. The video searching methodology check the content of index attached with each video to ensure that
video is matching with the searching keyword and its relevance ensured, based on the word count of searching keyword in video index.
Video captions are generated by the deep learning network model by combining global local (glocal) attention and context cascading
mechanisms using VIST-Visual Story Telling dataset. Video Index generator uses Wormhole algorithm, that ensure minimum worst-case
time for searching a key with a length of L. Video searching methodology extracts the video clip where the frames of interest lies from the
original huge sized source video. Hence, searcher can get and download a video clip instead of downloading entire video from the video
storage. This reduces the bandwidth requirement and time taken to download the videos.
关键词：Video Indexing; Video Searching; Visual Story Telling; Wormhole; glocal; VIST
