Video streaming services, including YouTube, Netflix, and Amazon, account for a large portion of global internet traffic. Therefore, the maintenance of high visual quality is very important to user satisfaction and overall Quality of Experience (QoE) for these users. However, fluctuating bandwidth often causes frame drops and motion discontinuities, resulting in visual artifacts and degraded QoE. Although existing video frame interpolation methods attempt to address these challenges, most remain limited in their ability to sustain performance under bandwidth-constrained environments. To overcome these limitations, IncepU-Net has been introduced. IncepU-Net is proposed as an enhanced U-Net architecture that integrates multi-scale inception modules and attention-gated skip connections for intelligent video frame interpolation in bandwidth-limited settings. IncepU-Net was trained and evaluated on the Vimeo90K and UCF101 datasets. The model at the testing stage achieved 35.19 dB PSNR and 0.958 SSIM using Vimeo90K, and 30.08 dB PSNR and 0.933 SSIM on UCF101. IncepU-Net surpasses several state-of-the-art interpolation methods. These results demonstrate that IncepU-Net effectively interpolates missing frames, mitigates playback stutter, and maintains motion continuity, thereby establishing it as a robust and practical solution for improving QoE in real-world video streaming applications, bridging the gap between restricted network delivery and perceptually meaningful frame synthesis.
Abbood et al. (Wed,) studied this question.