FastSHADE: Fast Self-augmented Hierarchical Asymmetric Denoising for Efficient inference on mobile devices

TL;DR

FastSHADE introduces a lightweight, efficient neural network architecture for real-time image denoising on mobile devices, balancing speed and quality through novel modules and augmentation strategies.

Contribution

The paper proposes FastSHADE, a novel mobile-friendly denoising network with unique modules and augmentation techniques, achieving state-of-the-art results in real-time mobile image restoration.

Findings

FastSHADE-M runs under 50 ms latency on Adreno 840 GPU.

FastSHADE-XL achieves 37.94 dB PSNR, setting new state-of-the-art.

The model family offers a scalable speed-fidelity trade-off.

Abstract

Real-time image denoising is essential for modern mobile photography but remains challenging due to the strict latency and power constraints of edge devices. This paper presents FastSHADE (Fast Self-augmented Hierarchical Asymmetric Denoising), a lightweight U-Net-style network tailored for real-time, high-fidelity restoration on mobile GPUs. Our method features a multi-stage architecture incorporating a novel Asymmetric Frequency Denoising Block (AFDB) that decouples spatial structure extraction from high-frequency noise suppression to maximize efficiency, and a Spatially Gated Upsampler (SGU) that optimizes high-resolution skip connection fusion. To address generalization, we introduce an efficient Noise Shifting Self-Augmentation strategy that enhances data diversity without inducing domain shifts. Evaluations on the MAI2021 benchmark demonstrate that our scalable model family establishes a highly efficient speed-fidelity trade-off. Our base FastSHADE-M variant maintains real-time latency (<50 ms on an Adreno 840 GPU) while preserving structural integrity, and our scaled-up FastSHADE-XL establishes a new state-of-the-art for overall image quality, achieving 37.94 dB PSNR.