Image defocus is inherent in the physics of image formation caused by the optical aberration of lenses, providing plentiful information on image quality. Unfortunately, the existing quality enhancement approaches for compressed images neglect the inherent characteristic of defocus, resulting in inferior performance. This paper finds that in compressed images, the significantly defocused regions are with better compression quality and two regions with different defocus values possess diverse texture patterns. These findings motivate our defocus-aware quality enhancement (DAQE) approach. Specifically, we propose a novel dynamic region-based deep learning architecture of the DAQE approach, which considers the region-wise defocus difference of compressed images in two aspects. (1) The DAQE approach employs fewer computational resources to enhance the quality of significantly defocused regions, while more resources on enhancing the quality of other regions; (2) The DAQE approach learns to separately enhance diverse texture patterns for the regions with different defocus values, such that texture-wise one-on-one enhancement can be achieved. Extensive experiments validate the superiority of our DAQE approach in terms of quality enhancement and resource-saving, compared with other state-of-the-art approaches.