The spatial pattern of vegetation changes during ecological restoration, and these changes are affected by the process of restoration. The objective of this study was to integrate the pattern and mechanism of forest restoration in the Dinghushan Nature Reserve (DNR), Guangdong, China, based on data from remote sensing and long-term field observations. We studied the pattern dynamics of three main forest types and their underlying mechanisms during restoration following a multiscale, hierarchical patch dynamics framework that integrates population, community, and landscape processes. Remote sensing data were used to determine the changes in landscape pattern during different periods of forest restoration from 1978 to 2006. At the landscape scale, the number, area, and perimeter of the needle/broad-leaved mixed forest (MF) and the evergreen broad-leaved forest (BF) increased, whereas those of the tropical needle-leaved forest (NF) decreased during succession. Our analysis based on long-term field observations indicated that the change rate of NF was lower than that of MF during 1981–1996, but became much higher during 1996–2007. The rate of change in landscape pattern and the progression of succession stages were consistent with each other. Our results also showed that species regeneration and community succession are the biological basis of forest landscape dynamics during vegetation restoration. Landscape pattern analysis allowed us to show “what” happened during vegetation restoration and “where,” and population and community analysis indicated “why” and “how” it happened.