Textureless building surfaces composed of homogenized pixels could lead to failure of photometric consistency. However, textureless regions widely present in artificial scenes usually exhibit strong planarity enabling depth estimation of textureless regions with planar priors. However, existing methods for generating planar priors suffer from over-segmentation of large planes with textureless regions, which indicates that planarity is not fully exploited. In this study, we propose a novel generation method of planar prior by combining mean-shift clustering and superpixel segmentation. The planarity is fully utilized given preferential generation of planar priors for large planes with textureless regions in artificial scenes. Finally, a probabilistic graphical model is used to adopt the planar priors and smoothing constraints into depth estimation process. The image gradient is used as a criterion of the degree of texture to adaptively adjust the weights of different constraints. Experimental results on the benchmark dataset ETH3D, UDD5, and SenseFly demonstrate that the proposed method can effectively recover the depth information of textureless regions in high-resolution images to obtain highly complete three-dimensional (3-D) models of artificial scenes.

Multiview stereo (MVS) achieves efficient 3D reconstruction on Lambertian surfaces and strongly textured regions. However, the reconstruction of weakly textured regions, especially planar surfaces in weakly textured regions, still faces significant challenges due to the fuzzy matching problem of photometric consistency. In this paper, we propose a multiview stereo for recovering planar surfaces guided by confidence calculations, resulting in the construction of large-scale 3D models for high-resolution image scenes. Specifically, a confidence calculation method is proposed to express the reliability degree of plane hypothesis. It consists of multiview consistency and patch consistency, which characterize global contextual information and local spatial variation, respectively. Based on the confidence of plane hypothesis, the proposed plane supplementation generates new reliable plane hypotheses. The new planes are embedded in the confidence-driven depth estimation. In addition, an adaptive depth fusion approach is proposed to allow regions with insufficient visibility to be effectively fused into the dense point clouds. The experimental results illustrate that the proposed method can lead to a 3D model with competitive completeness and high accuracy compared with state-of-the-art methods.