JND-Guided Light-Weight Neural Pre-Filter for Perceptual Image Coding

2025-10-12

AI Summary

To address the high computational cost of Just-Noticeable-Difference (JND)-guided pre-filtering and the lack of a unified evaluation benchmark, this paper proposes a lightweight frequency-domain JND pre-filtering method. First, we introduce FJNDF-PyTorch—the first open-source, standardized benchmark for evaluating frequency-domain JND pre-filtering. Second, we design a differentiable, end-to-end CNN-based framework that jointly optimizes human visual system modeling and image coding, achieving substantial bitrate reduction while preserving perceptual quality. Third, our method enables highly efficient inference, requiring only 7.15 GFLOPs for 1080p images—just 14.1% of the computational cost of recent lightweight models. Extensive experiments across multiple datasets and encoders demonstrate state-of-the-art perceptual compression performance.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding
• Machine Learning: Learning on the Edge & Model Compression

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Just Noticeable Distortion (JND)-guided pre-filter is a promising technique for improving the perceptual compression efficiency of image coding. However, existing methods are often computationally expensive, and the field lacks standardized benchmarks for fair comparison. To address these challenges, this paper introduces a twofold contribution. First, we develop and open-source FJNDF-Pytorch, a unified benchmark for frequency-domain JND-Guided pre-filters. Second, leveraging this platform, we propose a complete learning framework for a novel, lightweight Convolutional Neural Network (CNN). Experimental results demonstrate that our proposed method achieves state-of-the-art compression efficiency, consistently outperforming competitors across multiple datasets and encoders. In terms of computational cost, our model is exceptionally lightweight, requiring only 7.15 GFLOPs to process a 1080p image, which is merely 14.1% of the cost of recent lightweight network. Our work presents a robust, state-of-the-art solution that excels in both performance and efficiency, supported by a reproducible research platform. The open-source implementation is available at https://github.com/viplab-fudan/FJNDF-Pytorch.

Problem

• Developing lightweight JND-guided pre-filters for perceptual image compression
• Addressing computational expense in existing perceptual coding methods
• Establishing standardized benchmarks for fair comparison of pre-filters

Innovation

• Lightweight CNN for perceptual image pre-filtering
• Unified JND benchmark for fair method comparison
• Reduced computational cost with 7.15 GFLOPs requirement