A content-aware video adaptive transmission method based on deep learning network

Abstract

The invention discloses a content-aware video adaptive transmission method based on a deep learning network. The method is a new video transmission framework, and the computing power of the client canbe effectively utilized. According to video types, specific content-aware processing is carried out at the server side, core effective information is extracted for coding processing. Definition reconstruction is carried out on the downloaded video in the server at the client, so that a relatively low-quality video can reach relatively good video quality through processing of a deep learning network, and the method can effectively reduce dependence of a video stream on a bandwidth and improve watching experience of a user.

Description

technical field [0001] The invention belongs to the technical field of mobile network transmission, and relates to a server-side video deployment method and processing method in the network transmission process, as well as a super-resolution reconstruction processing process of a requested video, and in particular to a content-aware video auto-sensing based on a deep learning network. Adapt to the transfer method. Background technique [0002] With the continuous development of mobile Internet technology and the popularization of smart terminal devices, video services have gradually played an increasingly important role in Internet services. However, traditional video transmission mechanisms rely heavily on the state of network bandwidth. Therefore, the user's viewing experience will be greatly affected by the state of the network bandwidth. However, the proposal of technologies such as HTTP Adaptive Streaming (HAS, HTTP Adaptive Streaming), which can effectively reduce th...

AI Technical Summary

Problems solved by technology

[0006] In order to overcome the shortcomings of the above-mentioned prior art, the object of the present invention is to provide a content-aware video adaptive transmission method based on a deep learning network, which utilizes related applications of the deep learning network to process the server video through a convolutional neural network, Take effective information; re-process the video definition on the client side to improve the user experience of video applications in the mobile network; finally, through different operations on the client side and server side, solve the user viewing experience under the condition of poor bandwidth resources The problem

Images