Seeing enough: non-reference perceptual resolution selection for power-efficient client-side rendering
2026-04-09 • Graphics
Graphics
AI summaryⓘ
The authors developed a way to automatically find the lowest video resolution that still looks just as good to people as the highest quality version, without needing a reference video to compare against. This method uses knowledge about how human vision works over space and time, allowing devices to save power by not rendering unnecessarily high-resolution video. Their system is flexible, works with various video formats, and needs only small changes to current setups. It was trained on many videos labeled with a complex, full-reference quality metric but can predict quality using only the video itself. This approach helps improve visual quality while reducing computing costs on devices.
perceptual video qualitynon-reference metricspatio-temporal limitsclient-side renderingpower efficiencycodec-agnosticfull-reference metricmachine learningvideo resolutioncomputational cost
Authors
Yaru Liu, Dayllon Vinícius Xavier Lemos, Ali Bozorgian, Chengxi Zeng, Alexander Hepburn, Arnau Raventos
Abstract
Many client-side applications, especially games, render video at high resolution and frame rate on power-constrained devices, even when users perceive little or no benefit from all those extra pixels. Existing perceptual video quality metrics can indicate when a lower resolution is "good enough", but they are full-reference and computationally expensive, making them impractical for real-world applications and deployment on-device. In this work, we leverage the spatio-temporal limits of the human visual system and propose a non-reference method that predicts, from the rendered video alone, the lowest resolution that remains perceptually indistinguishable from the best available option, enabling power-efficient client-side rendering. Our approach is codec-agnostic and requires only minimal modifications to existing infrastructure. The network is trained on a large dataset of rendered content labeled with a full-reference perceptual video quality metric. The prediction significantly enhances perceptual quality while substantially reducing computational costs, suggesting a practical path toward perception-guided, power-efficient client-side rendering.