Dual-Level Precision Edges Guided Multi-View Stereo with Accurate Planarization

The reconstruction of low-textured areas is a prominent research focus inmulti-view stereo (MVS). In recent years, traditional MVS methods haveperformed exceptionally well in reconstructing low-textured areas byconstructing plane models. However, these methods often encounter issues suchas crossing object boundaries and limited perception ranges, which underminethe robustness of plane model construction. Building on previous work(APD-MVS), we propose the DPE-MVS method. By introducing dual-level precisionedge information, including fine and coarse edges, we enhance the robustness ofplane model construction, thereby improving reconstruction accuracy inlow-textured areas. Furthermore, by leveraging edge information, we refine thesampling strategy in conventional PatchMatch MVS and propose an adaptive patchsize adjustment approach to optimize matching cost calculation in bothstochastic and low-textured areas. This additional use of edge informationallows for more precise and robust matching. Our method achievesstate-of-the-art performance on the ETH3D and Tanks & Temples benchmarks.Notably, our method outperforms all published methods on the ETH3D benchmark.