Multimode fiber (MMF) contains rich temporal and spatial information because of the mode characteristics. As a novel optical fiber imaging method, MMF imaging based on temporal-spatial information extraction has the advantages of small device size, high resolution, large information capacity, and minimal invasion and has the potential to become a new generation of high-resolution and low-invasive endoscope. This review summarizes the basic methods and related progress of MMF imaging and introduces the related work of combining machine learning with MMF imaging. In addition, for the practical application of MMF imaging, we discuss the methods and related progress of MMF imaging under dynamic perturbation and the limitation of MMF imaging performance and quality. Finally, MMF imaging is summarized and prospected.