Q-Refine is a novel quality-aware refiner designed to enhance the quality of AI-Generated Images (AIGIs). It addresses the challenge of uniformly refining AIGIs of varying qualities, which can lead to negative optimization in high-quality regions and insufficient enhancement in low-quality regions. Q-Refine introduces an Image Quality Assessment (IQA) metric to guide the refining process, inspired by the Human Visual System (HVS). The method consists of three adaptive pipelines: (1) Gaussian Noise, which adds noise to low-quality (LQ) regions to trigger the denoising mechanism; (2) Mask Inpainting, which uses a smoothed quality map to modify LQ and MQ regions while retaining HQ regions; and (3) Global Enhancement, which fine-tunes low-level attributes to improve overall image quality. Experimental results on three AIGI quality databases (AGIQA-3K, AGIQA-1K, and AIGCIQA) demonstrate that Q-Refine effectively improves the quality of AIGIs at both fidelity and aesthetic levels, outperforming existing refiners in various scenarios.Q-Refine is a novel quality-aware refiner designed to enhance the quality of AI-Generated Images (AIGIs). It addresses the challenge of uniformly refining AIGIs of varying qualities, which can lead to negative optimization in high-quality regions and insufficient enhancement in low-quality regions. Q-Refine introduces an Image Quality Assessment (IQA) metric to guide the refining process, inspired by the Human Visual System (HVS). The method consists of three adaptive pipelines: (1) Gaussian Noise, which adds noise to low-quality (LQ) regions to trigger the denoising mechanism; (2) Mask Inpainting, which uses a smoothed quality map to modify LQ and MQ regions while retaining HQ regions; and (3) Global Enhancement, which fine-tunes low-level attributes to improve overall image quality. Experimental results on three AIGI quality databases (AGIQA-3K, AGIQA-1K, and AIGCIQA) demonstrate that Q-Refine effectively improves the quality of AIGIs at both fidelity and aesthetic levels, outperforming existing refiners in various scenarios.