Overview of Real-Time Rendering Super-Resolution and Frame Generation

As real-time rendering incorporates increasingly complex content and higher display specifications, graphics devices are approaching performance bottlenecks. In this context, super-resolution (SR) and frame generation (FG) solutions offer the possibility of achieving visual effects comparable to native rendering while operating at low resolutions and frame rates. Different from general SR and FG technologies, SR and FG for real-time rendering face stricter performance requirements while accessing more comprehensive auxiliary information from rendering engines, thus playing a crucial role in fields such as virtual reality (VR) and video games. This paper provides a review of SR and FG technologies for real-time rendering, categorizes relevant methods into three types, and conducts an analysis from multiple perspectives. First, the basic definitions and core concepts of SR and FG are introduced. Subsequently, related methods are classified into three categories—SR-only, FG-only, and joint SR-FG—then analyzed and compared. Finally, the research progress of SR and FG technologies in real-time rendering is summarized, and potential future research directions are identified, including scene-aware SR and FG, hardware implementation of relevant algorithms, and parallelization of SR/FG with normal rendering.