Colored point cloud reconstruction based on image inpainting

Base and floor maps are required for outdoor and indoor applications. In unknown environments, laser scanning is a more adequate approach. However, in laser scanning, when there are large missing and occluded areas, it is not easy to generate complete base and floor maps automatically. Thus, we focused on an image inpainting for colored point cloud rendering. Our approach was aimed at improving the quality of a digital elevation model (DEM) and ortho images generated from colored point cloud data. Our inpainting procedure consisted of base patch extraction, matching point detection, and image reconstruction steps. These steps were iterated for the reconstruction of a natural image. We selected two places for our study areas, an indoor and an outdoor environment. We have developed an inpainting approach for colored DEM generation using a colored point cloud. Moreover, we conducted two experiments using a terrestrial laser scanner and an indoor mobile mapping system. Through our experiments, we have confirmed that our approach could reconstruct natural textures and colored DEM from colored point cloud data with missing areas.