Seeing enough: non-reference perceptual resolution selection for power-efficient client-side rendering

2026-04-09

AI Summary

This work addresses the redundant energy consumption caused by high-resolution rendering on power-constrained devices and the impractical computational overhead of existing full-reference perceptual quality assessment methods. The authors propose a lightweight, no-reference approach for perceptual resolution selection that leverages the spatiotemporal limits of human visual perception to predict the lowest resolution at which users cannot discern visual differences from the original render, thereby enabling perceptually lossless yet energy-efficient rendering. The method is codec-agnostic and requires only minimal modifications to existing rendering pipelines. Trained on a large-scale rendered video dataset using full-reference metrics as supervision, the underlying neural network supports efficient on-device inference. Experimental results demonstrate that the proposed solution substantially reduces computational cost while preserving high perceptual quality, offering a practical strategy for energy-aware client-side rendering.

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.

Problem

• perceptual quality
• non-reference
• resolution selection
• power-efficient rendering
• client-side

Innovation

• non-reference
• perceptual quality
• resolution selection
• power-efficient rendering
• human visual system