Assessing the Impact of Color Adjustments on Photogrammetry-Based 3D Surface Reconstruction

Abstract

Photogrammetry has evolved as an important technique for generating point clouds from 2D images especially in the absence of direct 3D sensors for acquiring 3D points. It is seen in several image processing applications that the quality of the result depends on several parameters including luminosity, color saturation, and hue. While the technical workflow of the photogrammetry is well defined but the coloring aspects of the image and its effect on the quality of the point cloud generation has been less explored. This paper studies the impact of color adjustments including contrast, brightness, and gamma variation on 3D point cloud reconstruction capability. The result shows that higher contrast above 40% gradually reduces the reconstruction quality in terms of points on the surface, defining vertex set and faces. Similar observations are also true for brightness values. On negative side beyond -15% tonal adjustments, the results degraded faster. Other interesting results and its potential applications have also been cited in this paper. RMSE observations on resulting dense point cloud, faces and control vertices exhibit substantial loss of quality of reconstructed surface for higher values of brightness and contrast. Several future pointers for improving the reconstruction quality have also been proposed.