Monocular depth estimation (MDE)—the task of predicting dense scene depth from a single RGB image—has emerged as a fundamental computer vision problem with widespread applications in autonomous driving

robotic navigation

augmented reality

and medical imaging. Unlike stereo or multi-view approaches

MDE must resolve inherent scale ambiguity by learning implicit geometric cues from texture

structure

and semantic context

making it a uniquely challenging ill-posed problem. Particular attention is given to emerging challenges: robustness to domain shift

performance in low-texture and dynamic scenes

scale ambiguity resolution

and the trade-offs between accuracy and inference efficiency for real-world deployment. Finally

we synthesize current research bottlenecks and outline promising future directions

including foundation models for depth

geometric-aware generative approaches

and robust multi-task learning frameworks. This survey aims to provide researchers and practitioners with a structured understanding of the field's evolution and a roadmap for future investigations.