A review of reinforcement learning in image driven applications

Reinforcement Learning (RL) has evolved into a fundamental paradigm within image-related research, propelled by significant progress in machine learning, computer vision, robotics, and medical imaging. In accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines, this review systematically investigates the progression of RL applications in image analysis between 2011 and 2025. Initial studies predominantly addressed foundational aspects, including policy optimization and visual task learning within controlled or simulated environments. In contrast, contemporary research exhibits a marked transition toward broader thematic areas, methodological innovations, and applied implementations across diverse real-world contexts. Furthermore, this review delineates persisting challenges and identifies emerging research trajectories, offering critical insights and strategic guidance for advancing RL-based approaches in image analysis.