Real-Time Video PLC Using In-Painting

Video packet loss during network transmission can lead to visible artifacts, freezing, and interruptions in video playback. Packet loss concealment techniques aim to mitigate these effects by concealing missing packets and enhancing the viewing experience. This work focuses on real-time concealment...

Full description

Saved in:
Bibliographic Details
Main Authors: Katiyar, Rajani, Chakraborty, Prasenjit, Surana, Ravi, Holla, Ravishankar, Sanjana, Sanka, Acharya, Sathvik, B, Sonia Singh, Agrawal, Yash
Format: Conference Proceeding
Language:English
Subjects:
Android Application
Deep learning
Frame Prediction
Packet loss
Packet Loss Concealment
Predictive models
Quality assessment
Real-time systems
Streaming media
Video Transmission
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Video packet loss during network transmission can lead to visible artifacts, freezing, and interruptions in video playback. Packet loss concealment techniques aim to mitigate these effects by concealing missing packets and enhancing the viewing experience. This work focuses on real-time concealment of packet loss using a deep learning model deployed within an Android application. The objective is to utilize frame prediction to conceal packet loss and improve video quality. This work presents a novel approach to real-time concealment of video packet loss, addressing the absence of existing solutions for the real time concealment. According to need of the work dataset was created, meticulously curated to facilitate the training and testing of a deep learning model. A deep learning model is integrated into an Android application for real-time packet loss detection. Frame prediction conceals lost packets by generating predicted frames, reducing the impact of packet loss for uninterrupted viewing. The implemented model achieves efficient packet loss detection and frame prediction, with an average prediction time of under 30ms per frame. This rapid prediction time contributes to minimal latency and reduced visual artifacts during packet loss concealment. In response to the challenges inherent in real-time Packet Loss Concealment (PLC), this work presents a solution that prioritizes the enhancement of video quality, all while meticulously managing stringent low inference time requirements, minimizing battery consumption, and seamlessly handling the real-time processing of video data, most of which have not been adequately addressed in existing literature.
ISSN:2155-2509
DOI:10.1109/COMSNETS59351.2024.10427265