| Description | this study proposes a novel Monocular-Guided Edge-Enhanced Deformation MVS (MED-MVS) approach that deeply integrates high-precision monocular depth priors and edge-aware refinement mechanisms into the entire MVS pipeline. Specifically, MoGe-2 is adopted to generate metric-scale monocular depth priors from raw input images; a monocular-depth-aware deformable propagation strategy is designed to refine multi-scale depth hypotheses; and monocular-guided depth filtering is introduced between pyramid level processing steps. The proposed ME-MVS achieves reliable depth range guidance by selecting valid guide points from monocular depth priors, makes up for the insufficiency of photometric and geometric consistency in cost computation for high-reflection and fine-detail regions, and eliminates outliers via inter-level monocular-guided filtering to ensure the reliability of iterative optimization. |