A technique to ensure color fidelity in automatic photogrammetry

In this conference we presented in 2013 a new process and some solutions for color acquisition, management, rendering and assessment in Architectural Heritage (AH) 3D models construction from reality-based data based on photographs, with the aim to illustrate easy, low-cost and rapid procedures that produce high visual accuracy of the image/model while being accessible to non-specialized users and unskilled operators, typically Heritage architects. The 2013 solution was successfully tested on the digitization as 3D models of 10 kilometers of buildings with porticoes but its use demonstrated some bottleneck resulting mainly by the lack of automation in many steps. In the last years the image-based pipeline for 3D reconstruction purposes has received large interest leading to fully automated methodologies able to process large image datasets and deliver 3D products with a level of detail and precision variable according to the applications. Most part of developed solutions allows to improve our pipeline, but none of them provides the solution to the problem of accurate color reproduction using automated processes. In this paper we present an automated solution to increase the processing performances of the two central steps of the photogrammetric pipeline (camera orientation & bundle adjustment) and simultaneously radiometrically improve the quality of an image datasets. Our solution aim to achieve a robust automatic color balance and exposure equalization to ensure (i) faithful color appearance of a digitized artifact and (ii) consistency of the color-to-gray conversion. A correct color balance allows to minimize the typical problem of incorrectly detected areas (e.g. different luminance value for the same color and/or isoluminant colors) that strongly appear in case of surfaces of the same color or colors with the same luminance value. Therefore, the color balance procedure aims to produce radiometrically-calibrated images ensuring the consistency of surface colors in all the images (i.e. as much as possible similar RGB values for homologous pixels). Starting from captured RAW images our workflow includes: exposure compensation, adaptive histogram equalization and color balance. Between the two general approaches (color characterization vs spectral sensitivities based on color targets) we adopted this last technique that uses a set of differently colored samples measured with a spectrophotometer. A software exploiting a series of Xrite ColorChecker Classic was developed and a series of trials were put in place to demonstrate the efficiency of our automatic process in different case of use. Color accuracy (‘color characterization’) was evaluated and expressed.