Q-Refine is a quality-aware refiner for AI-generated images (AIGIs) that addresses the challenge of uniformly refining AIGIs of varying qualities. Traditional refiners often fail to optimize low-quality images without negatively affecting high-quality ones. Q-Refine introduces the Human Visual System (HVS) preference and uses Image Quality Assessment (IQA) metrics to guide the refining process. It employs three adaptive pipelines for low/medium/high quality regions, enabling targeted optimization. The framework includes a quality pre-processing module and three refining pipelines inspired by predicted quality. The first pipeline adds noise to low-quality regions to help achieve global optimization. The second uses a mask to retain high-quality areas while modifying lower quality regions. The third enhances the overall image quality. Experimental results show that Q-Refine effectively improves the quality of AIGIs across different levels, achieving state-of-the-art performance on multiple databases. It outperforms existing refiners in terms of both fidelity and aesthetic quality, demonstrating its versatility and effectiveness in refining AIGIs without negatively impacting high-quality regions. The method is validated on three AIGI quality databases, showing significant improvements in image quality metrics. Q-Refine's adaptive pipelines and IQA integration make it a general refiner for AIGIs, enhancing their visual quality and usability.Q-Refine is a quality-aware refiner for AI-generated images (AIGIs) that addresses the challenge of uniformly refining AIGIs of varying qualities. Traditional refiners often fail to optimize low-quality images without negatively affecting high-quality ones. Q-Refine introduces the Human Visual System (HVS) preference and uses Image Quality Assessment (IQA) metrics to guide the refining process. It employs three adaptive pipelines for low/medium/high quality regions, enabling targeted optimization. The framework includes a quality pre-processing module and three refining pipelines inspired by predicted quality. The first pipeline adds noise to low-quality regions to help achieve global optimization. The second uses a mask to retain high-quality areas while modifying lower quality regions. The third enhances the overall image quality. Experimental results show that Q-Refine effectively improves the quality of AIGIs across different levels, achieving state-of-the-art performance on multiple databases. It outperforms existing refiners in terms of both fidelity and aesthetic quality, demonstrating its versatility and effectiveness in refining AIGIs without negatively impacting high-quality regions. The method is validated on three AIGI quality databases, showing significant improvements in image quality metrics. Q-Refine's adaptive pipelines and IQA integration make it a general refiner for AIGIs, enhancing their visual quality and usability.