The Technology Of Smart Quick Browsing Based On Key-frame Extraction

With the rapid development of Internet and digital cameras technology,the storage of video data continues to rise sharply.Since the diverse and abundant video resources have sprung up,there has been a growing demand for efficient methods to understand the video quickly.It’s a feasible scheme that dragging the seeking bar manually back and forth to review the video.However,it requires a significant dedication of time and effort when implementing the operation and looking at the screen fixedly.Especially in long surveillance videos with discontinuous stable activities,users don’t have enough time to browse them.How to understand the content in a limited time has become a hot topic.This paper proposed an adaptive fast playback framework which is intended for meeting the requirement of hash time terms.This framework is based on the extraction of key frame where multi-features are used to maintain a stable playback“pace”.Firstly,Jenson noise-based difference(JSND)measure is proposed to describe the difference of the temporal adjacent frames,on the basic of which a visual interest learning model is used to study the users browsing behavior to speculate the hobbies and interests.Thus,this system support a satisfactory playback rate and content and at the same time the uninterested content will be fast skipped.A non-linear frame skipping strategy is specially designed for two different video types and then an image fusion algorithm is applied for a long video with an arbitrary frame-rate.Finally,a visual browsing interface is realized with the metaphor of car driving to help users to understand the main content and appreciate the details.The experience shows that the design of visual system can help users quickly scan the content,keep the video playback fluency and void the miss of main interested parts,winning a wonderful experience.