Video transmission acceleration method, transmitting end, receiving end and storage medium

By employing a super-resolution deep neural network method involving inter-frame partitioning and retraining in video transmission, the problem of insufficient computing power in the cloud and on-device is solved, enabling faster and more efficient high-definition video transmission on mobile devices.

CN115914736BActive Publication Date: 2026-06-16SHANGHAI JIAOTONG UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SHANGHAI JIAOTONG UNIV
Filing Date
2022-09-27
Publication Date
2026-06-16

AI Technical Summary

⚠Technical Problem

Existing video transmission acceleration methods assume that the cloud and the device have unlimited computing power and high-resolution video resources, resulting in insufficient processing speed when playing videos on mobile devices, and failing to effectively reduce transmission time and bandwidth consumption.

⚗Method used

The video stream is divided into two parts using an inter-frame partitioning algorithm. One part is processed by a super-resolution deep neural network and encoded and compressed at the transmitting end, while the other part is directly compressed and transmitted to the receiving end for processing. FFmpeg is used for H265 lossy encoding, and a retrained super-resolution deep neural network is used to improve the loss caused by lossy compression.

🎯Benefits of technology

By optimizing the inter-frame partitioning ratio, fully utilizing the computing resources of both the transmitter and receiver, video transmission time is significantly shortened, bandwidth consumption is saved, and image quality is improved through retraining, enabling faster acquisition of high-definition video.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115914736B_ABST
    Figure CN115914736B_ABST
Patent Text Reader

Abstract

The application provides a video transmission acceleration method, a transmitting end, a receiving end and a storage medium, and the method comprises the following steps: dividing a video stream into a first video frame stream and a second video frame stream based on an interframe division algorithm; processing the first video frame stream based on a super-resolution deep neural network to obtain a first image stream, performing encoding and compression processing on the first image stream to obtain a first data stream, and sending the first data stream to a receiving end; performing encoding and compression processing on the second video frame stream to obtain a second data stream, and sending the second data stream to the receiving end. According to the current network transmission environment and the hardware resource conditions of the transmission end, an optimal interframe division ratio can be selected, the bidirectional operation resources of the video transmitting end and the receiving end are fully utilized to realize the acceleration of the video transmission process, the bandwidth consumption is greatly saved, and the loss caused by the lossy compression transmission of the video is improved by using the deep neural network retraining method.
Need to check novelty before this filing date? Find Prior Art