All aforementioned sources of error influence the reconstruction of the final model. But there is a problem in determining the geometric accuracy for reconstructions of real scenes perse . This is due to the f act that one cannot locate
points on the reconstructed model for which the real-world coordinates are known without considering its texture (e.g. by using special markers that are located in the reconstructed model). This implies that the (mis-)alignment of textures on the model will bias the accuracy analysis of reconstructed 3D models in general. Besides that, the perceived detail level strongly depends on the resolution of the model's texture. So even if images with reduced resolution were used for the computation we take the original images as textures. This enables the user to identify points of interest in the
model more reliably.