Abstract

Advancements in artificial intelligence have revolutionized real-time video processing, enabling enhanced visual quality for applications in surveillance, medical imaging, and entertainment. Traditional video enhancement methods often struggle with balancing computational efficiency and high-quality output, leading to degraded performance in real-time scenarios. The primary challenge lies in preserving details while reducing noise, motion artifacts, and frame inconsistencies, particularly in low-resolution and fast-motion videos. This study introduces an AI-driven real-time video enhancement framework that integrates super-resolution techniques with optical flow-based motion estimation. The proposed method employs a deep learning-based Super-Resolution Generative Adversarial Network (SRGAN) to upscale video frames while maintaining texture fidelity. Additionally, an enhanced optical flow algorithm refines motion estimation, minimizing temporal inconsistencies and improving frame transitions. The combination of these techniques enables effective noise reduction, sharper details, and smooth motion handling, making the framework suitable for real-time applications. Experimental evaluations demonstrate that the proposed approach significantly improves peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) compared to existing methods. The system achieves real-time performance with minimal computational overhead, making it suitable for deployment in live broadcasting, telemedicine, and security surveillance. The results highlight the efficiency of integrating AI- based super-resolution with optical flow in achieving superior video clarity and motion coherence in real-time environments.

Authors

Sonia Victor Soans¹, T. Prabakaran², S. Vamshi Krushna³
University of Technology and Applied Sciences, Sultanate of Oman¹, Joginpally B.R. Engineering College, India², Samskruti College of Engineering and Technology, India³

Keywords

Real-Time Video Enhancement, Super-Resolution, Optical Flow, Ai- Driven Video Processing, Motion Estimation