Research On WebRTC Video Frame Rate Enhancement Algorithm And Design Implementation For Conference System

With the continuous emergence of various new technologies,WebRTC came into being.It is a browser-based multimedia real-time communication technology.Its emergence has opened a new era in the webpage era,but it does not stop at the web.WebRTC is a standardized abstraction of the terminal audio and video interface and the establishment of real-time communication,multimedia coding formats and standard specifications for transmission.As long as these standards are met,both the Web side and the mobile side can be applied.In addition to new technologies such as WebRTC,video post-processing technologies associated with WebRTC are also constantly being researched and updated.Among them,the frame rate improvement technology is a research hotspot in video post-processing technology.It has a wide range of applications in the digital media field.It can be used not only to improve the quality of video on the playback end,but also to balance the contradiction between massive video data and limited bandwidth resources.This thesis first studies the overall framework and core technologies of WebRTC,and studies the frame rate adjustment strategies that affect WebRTC video quality.Since a certain frame dropping strategy is adopted in WebRTC's real-time video technology,the frame rate improvement technology in video post-processing technology is studied.This thesis mainly researches and analyzes the traditional frame rate improvement algorithms,and at the same time researches and analyzes neural networks such as CNN,U-net,and Res Net.Aiming at the shortcomings of traditional frame rate improvement algorithms and combining the advantages of deep learning,this thesis proposes a frame rate improvement algorithm based on improved U-Net.This algorithm uses a combination of U-net and Res Net,and takes two adjacent frames of video as input,and outputs interpolation frames through U-net network encoding and decoding operations.The experimental results show that the video interpolation frames predicted by this deep learning-based frame rate improvement algorithm,this algorithm can effectively avoid the block effect of the traditional frame rate improvement algorithm and improve the sensory quality of the video.Finally,with the requirements of the project,this thesis designs and implements the WebRTC video conference system that based on Android and the research of WebRTC core technology.The system is mainly divided into client and server,and the server mainly meets the services such as streaming media data transmission,signaling service,and NAT penetration.For the transmission of streaming media data,a server based on the SFU communication architecture is mainly used,the signaling service is mainly implemented by using the Web Socket protocol,and NAT penetration mainly uses a server combining STUN and TURN.The client mainly includes a user registration and login module,a video conference module,an IM instant message module,and a live broadcast module.After the system is implemented,the system can basically meet the functional requirements of each module through functional test and performance test analysis of the system,and can guarantee normal installation,using,and uninstallation.