QuantDemoire: Quantization with Outlier Aware for Image Demoiréing

2025-10-05

AI Summary

Deploying deep learning demoiréing models on edge devices remains challenging due to significant performance degradation in existing post-training quantization (PTQ) methods, which neglect distribution outliers and cause representational degradation in smooth regions. To address this, we propose a lightweight PTQ framework tailored for demoiréing. Our method introduces an outlier-aware quantizer that jointly estimates sampling ranges and preserves extreme-weight values in FP16 format; and a frequency-domain-aware calibration strategy that employs frequency-weighted loss to emphasize low- and mid-frequency components while suppressing stripe artifacts induced by low-bit quantization. Using W4A4 mixed-precision quantization with lightweight fine-tuning, our approach substantially reduces computational overhead and model size. Quantitative evaluation shows a PSNR improvement of over 4 dB compared to state-of-the-art PTQ baselines, while preserving high-fidelity demoiréing performance.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Demoiréing aims to remove moiré artifacts that often occur in images. While recent deep learning-based methods have achieved promising results, they typically require substantial computational resources, limiting their deployment on edge devices. Model quantization offers a compelling solution. However, directly applying existing quantization methods to demoiréing models introduces severe performance degradation. The main reasons are distribution outliers and weakened representations in smooth regions. To address these issues, we propose QuantDemoire, a post-training quantization framework tailored to demoiréing. It contains two key components. **First}, we introduce an outlier-aware quantizer to reduce errors from outliers. It uses sampling-based range estimation to reduce activation outliers, and keeps a few extreme weights in FP16 with negligible cost. **Second**, we design a frequency-aware calibration strategy. It emphasizes low- and mid-frequency components during fine-tuning, which mitigates banding artifacts caused by low-bit quantization. Extensive experiments validate that our QuantDemoire achieves large reductions in parameters and computation while maintaining quality. Meanwhile, it outperforms existing quantization methods by over **4 dB** on W4A4. Code is released at: https://github.com/zhengchen1999/QuantDemoire.

Problem

• Reducing computational demands for image demoiréing on edge devices
• Addressing performance loss from outliers in quantized demoiréing models
• Mitigating banding artifacts in low-bit quantization of smooth regions

Innovation

• Outlier-aware quantizer reduces activation and weight errors
• Frequency-aware calibration emphasizes low and mid frequencies
• Post-training quantization framework for image demoireing models