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

Nikolay Falaleev

cs.CVApr 11, 2026

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

Authors: Nikolay Falaleev

Organizations: Fanis London, UK

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.

Figures & tables

Figure 1: Latency-Quality Pareto Frontier. The FastSHADE model family establishes a highly efficient trade-off between on-device inference time (measured in FP16 on a Qualcomm Adreno 840 GPU) and pixel-level restoration fidelity (PSNR). While ultra-lightweight methods like TLG operate faster, they suffer a noticeable degradation in image quality. In contrast, our scalable architecture significantly expands the state-of-the-art efficiency frontier, pushing the boundaries of visual fidelity while offering highly competitive, low-latency variants for real-world mobile denoising.

Figure 3: Frequency domain analysis of the internal features within the AFDB Block, averaged over 50 validation images. (Left, Center) The 2D FFT magnitude spectra exhibit a broader energy distribution in the $5\times 5$ path compared to the $3\times 3$ path. (Right) In the normalized radially averaged power spectrum, the spatial path (blue) retains significant energy in the low-to-mid spatial frequencies, indicating the preservation of structural details and edges. In contrast, the HF noise path (red) has a flat spectral response characteristic of isotropic, unstructured noise extraction.

Figure 4: Qualitative results of FastSHADE model predictions.

Method	SSIM $\uparrow$	PSNR $\uparrow$	$\mathbf{t}$ (ms) $\downarrow$	Score $\uparrow$
TLG	0.9129	37.31	28.0	13.72
SplitterNet [ 6 ]	-	37.62	66.1	8.93
TeamPAK	0.9098	37.58	127.0	4.40
Z6	0.9001	36.96	69.3	3.41
IVCL	0.9110	37.28	1961	0.19
FastSHADE-M	0.9136	37.52	48.1	10.68

Table 1: Performance metrics of FastSHADE compared to baselines and other methods. SSIM and PSNR values are reported for the MAI2021 test subset. Runtime,

t

, is reported for the Adreno 840. SSIM for SplitterNet was not reported because the test subset is not public.

Downscaling Layer	PSNR(dB) $\uparrow$	$\mathbf{t}$ (ms) $\downarrow$	Score $\uparrow$
Haar (frozen)	37.475	151	3.198
Haar (learnable)	37.492	151	3.275
$3\times 3$ Conv (stride 2)	37.581	141	3.967
$2\times 2$ Conv (stride 2)	37.580	136	4.108

Table 2: Ablation study on spatial downscaling strategies. Replacing the Haar wavelet transform and projection layer with a single learnable

2\times 2

strided convolution achieves the highest score.

Channel Split LF:HF	PSNR(dB) $\uparrow$	$\mathbf{t}$ (ms) $\downarrow$	Score $\uparrow$
1:0 (No Split)	36.56	144	3.773
1:1 (Symmetric)	37.58	142	3.934
1:3 (AFDB)	37.58	136	4.108

Table 3: Evaluation of channel splitting strategies in the AFDB. 1(LF):0(HF) routes all features through the

5\times 5

spatial branch; 1(LF):1(HF) represents a symmetric split; and 1(LF):3(HF) is our proposed asymmetric configuration (25%

5\times 5

spatial branch, 75%

3\times 3

noise). The asymmetric split matches the PSNR of the symmetric baseline but reduces inference time.

Upsampler Variant	PSNR(dB) $\uparrow$	$\mathbf{t}$ (ms) $\downarrow$	Score $\uparrow$
Concat + $1\times 1$ Conv	37.40	257	1.690
SE Attn.	37.37	145	2.863
CBAM Attn.	37.43	141	3.205
Tconv $2\times 2$ + Sigmoid	37.54	142	3.700
SGU (w/o $\text{Conv}_{3\times 3}^{dw}$ )	37.55	136	3.919
SGU	37.58	136	4.108

Table 4: Ablation study on skip connection fusion strategies in the upsampling block: compare the proposed SGU against standard U-Net concatenation, channel/spatial attention (SE and CBAM), and a transposed convolution baseline.

Data augmentation	PSNR(dB) $\uparrow$
Baseline	36.80
+ basic	37.58
+ basic + noise swap	37.51
+ basic + our	37.63

Table 5: Data augmentation techniques ablation results. Baseline means no augmentations; + basic adds random rotations, flips, and artificial Gaussian noise; (

\sigma=25

); + basic + noise swap includes basic augmentation and random noise profile swap between images in a batch; + basic + our adds the proposed Noise Shifting Self-Augmentation.

Image Denoising

Image denoising removes unwanted noise while preserving the edges, textures, and colors needed for faithful images.

See all 157 papers

Real Image Denoising with Knowledge Distillation for High-Performance Mobile NPUs
May 5, 2026Faraz Kayani, Sarmad Kayani, Asad Ahmed +2Efficient Neural Network Inference Knowledge Distillation
Computer Vision Lab, CAIDAS & IFI, University of Würzburg, Germany
LiteKD-Net: Lightweight Knowledge-Distilled Network for Mobile Image Denoising
Aug 6, 2026Zhou ZhiyiLightweight CNNs On-Device Inference
Global College Shanghai Jiaotong University Shanghai, China
Edge-Efficient Image Restoration: Transformer Distillation into State-Space Models
May 4, 2026Srinivas Soumitri Miriyala, Sowmya Vajrala, Sravanth Kodavanti +2Image Deraining Image Enhancement
Samsung Research Institute Bangalore, India

May 5, 2026J/K move · Enter open · S save

Real Image Denoising with Knowledge Distillation for High-Performance Mobile NPUs

Faraz Kayani, Sarmad Kayani, Asad Ahmed, Radu Timofte+1

Computer Vision Lab, CAIDAS & IFI, University of Würzburg, Germany

Efficient Neural Network Inference Knowledge Distillation On-Device Inference Image Denoising

Open paper AskPDF arXiv

While deep-learning-based image restoration has achieved unprecedented fidelity, deployment on mobile Neural Processing Units (NPUs) remains bottlenecked by operator incompatibility and memory-access overhead. We propose an NPU-aware hardware-algorithm co-design approach for real-world image denoising on mobile NPUs. Our approach employs a high-capacity teacher to supervise a lightweight student network specifically designed to leverage the tiled-memory architectures of modern mobile SoCs. By prioritizing NPU-native primitives -- standard 3x3 convolutions, ReLU activations, and nearest-neighbor upsampling -- and employing a progressive context expansion strategy (up to 1024x1024 crops), the model achieves 37.66 dB PSNR / 0.9278 SSIM on the validation benchmark and 37.58 dB PSNR / 0.9098 SSIM on the held-out test benchmark at full resolution (2432x3200) in the Mobile AI 2026 challenge. Following the official challenge rules, the inference runtime is measured under a standardized Full HD (1088x1920) protocol, where it runs in 34.0 ms on the MediaTek Dimensity 9500 and 46.1 ms on the Qualcomm Snapdragon 8 Elite NPU. We further reveal an "Inference Inversion" effect, where strict adherence to NPU-compatible operations enables dedicated NPU execution up to 3.88x faster than the integrated mobile GPU. The 1.96M-parameter student recovers 99.8% of the teacher's restoration quality via high-alpha knowledge distillation (alpha = 0.9), achieving a 21.2x parameter reduction while closing the PSNR gap from 1.63 dB to only 0.05 dB. These results establish hardware-aware distillation as an effective strategy for unifying high-fidelity denoising with practical deployment across diverse mobile NPU architectures. The proposed lightweight student model (LiteDenoiseNet) and its training statistics are provided in the NN Dataset, available at https://github.com/ABrain-One/NN-Dataset.

Abstract

Image Denoising

Explore similar work

Real Image Denoising with Knowledge Distillation for High-Performance Mobile NPUs

LiteKD-Net: Lightweight Knowledge-Distilled Network for Mobile Image Denoising

Edge-Efficient Image Restoration: Transformer Distillation into State-Space Models