Total Differential Photometric Mesh Refinement with Self-Adapted Mesh Denoising

Variational mesh refinement is a crucial step in multiview 3D reconstruction. Existing algorithms either focus on recovering details or suppressing noise. Approaches with consideration of both are lacking. To address this limitation, we proposed new variational method named total differential (TDR), which mainly included two improvements. First, the traditional partial-differential photo-consistency gradient used was replaced by total-differential gradient. With correlation between adjacent pixels, our can make achieve more effective convergence than approaches. Second, introduced bilateral normal filter novel self-adaptive denoising strategy into refinement. This maintains balance detail preservation and via zero-normalized cross-correlation (ZNCC) map. Various experiments demonstrated that superior to approaches accuracy effect. Moreover, compared generated open-source commercial software (Context Capture), meshes detailed, regular, smooth.