A Survey on Long-Range Stereo Depth Estimation in Outdoor Environments

Stereo depth estimation has been a central problem in computer vision for nearly five decades, evolving from purely geometric block-matching pipelines to large-scale data-driven deep architectures and, more recently, foundationmodel–based stereo systems. While the field has made remarkable progress on standard indoor and short-range outdoor benchmarks, long-range outdoor stereo depth estimation remains a substantially underexplored and uniquely challenging problem. The combination of small disparities at distance, low-texture surfaces such as roads and skies, repetitive architectural patterns, occlusions, illumination and weather variability, and the inverse depth– disparity relationship, jointly produces failure modes that neither purely traditional nor purely learned methods reliably address. This survey provides a focused, integrative review of stereo depth estimation methods relevant to long-range outdoor scenarios. We first formalize the geometric and statistical foundations of stereo correspondence under rectified configurations, then taxonomize traditional, deep-learning–based, and hybrid stereo methods according to their architectural primitives and their behavior under long-range conditions. We further review the principal benchmark datasets used to evaluate outdoor stereo systems, examine the evaluation metrics that are most informative for far-range performance, and analyze the open challenges that constrain progress in the field, including domain shift, adverse-weather robustness, sensitivity at small disparities, and computational deployment constraints. Throughout the paper we relate methodological choices to the geometric structure of the long-range outdoor stereo problem, with the aim of clarifying which design principles transfer across method families and which remain unsolved. The survey is intended both as an entry point for researchers new to long-range outdoor 2 stereo and as a structured reference for practitioners working on autonomous driving, robotics, and outdoor 3D reconstruction.