Search

Objects with textureless and specular surfaces are commonplace in manmade environments. Their reconstruction challenges optical 3-D metrology methods typically yielding incomplete and noisy results. Existing solutions, including structured light, shape from shading, and learning-based specular removal methods, place a high demand on projection patterns, the number of light sources and cameras, or training data. While requiring elaborate setups, these solutions are not general enough, and their success in handling these surfaces is only partial. To address this reconstruction challenge, we study in this paper the application of shape from polarization for the 3-D reconstruction of textureless objects. Using a single view and a known light source, polarization-based constraints can be expressed as a set of linear equations for the unknown depth. We then estimate depth directly as an optimization problem constrained by the linear polarization equations. Results demonstrate complete and accurate 3-D reconstructions of typical glossy featureless objects, suggesting that shape from polarization is a valuable strategy for generating dense 3-D surface models.

Three-dimensional modeling of heritage assets poses a challenge when high reflectance and featureless surfaces are involved. Because of the complex reflection characteristics, traditional methods, either passive or active, tend to generate noisy and incomplete reconstruction results. To address this, we introduce in this paper a shape-from-polarization system and modeling strategy for the 3-D reconstruction of heritage assets. As demonstrated, by analyzing polarization properties of reflected light, we can generate by a single image pixel-level-resolution normal and depth maps. We consider two system prototypes involving a point light source but different in the polarizing camera configurations. They provide solutions for different reconstruction requirements and acquisition conditions. Our reconstruction model is common however to both involving a closed set of simple computation steps. Our evaluations demonstrate how the proposed method significantly improves the 3-D surface model completeness and level of detail, showing its suitability and value to reach the goal of a high-resolution survey of heritage assets, where traditional approaches tend to fail.

The development of efficient strategies to document cultural heritage sites is an active research field. One promising avenue to address that need may be found in the use of portable laser scanners. Such scans provide a mapping-grade level of accuracy, yet their level of characterization is limited by the low resolution of the generated point cloud and by the relatively noisy measurements. In this paper we study methods to attenuate the noisy responses as a means to improve the data quality and highlight the underlying structure. Unlike the prevailing plane-fitting-based filtering approaches that tend to blur salient features, we consider the use of local structure properties in our denoising strategy. We use the normal-based bilateral filtering of point clouds as a platform, yet introduce new normal preservation concepts whose incorporation significantly improves the overall denoising process performance. Results demonstrate how our proposed solution outperforms the standard plane-filtering and the naive bilateral approach. The attenuation we achieve yields a more visually pleasing entity description as well as simplified processing of subsequent procedures, including feature extraction and semantic segmentation.