ATMOSPHERICALLY COMPENSATED SHAPE FROM SHADING ON THE MARTIAN SURFACE: TOWARDS THE PERFECT DIGITAL TERRAIN MODEL OF MARS
We have expanded our existing Shape and Albedo from Shading framework which has primarily been used to generate Digital Terrain Models (DTMs) of the Lunar Surface. The extension consists of an atmospheric model such that the approach can be applied to Mars which is covered by a thin atmosphere. The atmospheric model includes attenuation by the atmosphere, diffuse illumination of the surface and scattering from the atmosphere into the direction of the sensor with physically motivated parameters. To estimate the newly introduced atmospheric parameters without additional CRISM measurements, the radiance image and an initializing surface are used. The initial surface is derived from stereo images and serves two purposes. On the one hand, it is the height constraint of the SfS algorithm and on the other hand, it is used for estimating the atmospheric parameters. Relying on this estimation, the aforementioned Shape and Albedo from Shading method is carried out. The results show a considerable improvement compared to DTMs derived with stereo algorithms. The omnipresent stereo artifacts such as pixel locking and mismatches are smoothed out and small details are reconstructed convincingly. The procedure is then compared to the reconstruction without atmospheric compensation. Images in which shadows are present benefit from this method because shadows can now be described by the diffuse illumination of the surface. The reconstruction results indicate the viability of the approach since it can produce convincing DTMs compared to HiRISE ground truth.