I have been looking into the texture quality of textured 3D models.
The dataset I am basing the following observations on is of an urban environment, where each camera trigger event captures 5 x 24MP images, 1 x NADIR and 4 x 45 degree oblique in each direction. The NADIR images have a GSD of approximately 0.7 cm/px.
The best textured 3D model result I have been able to create is based on Depth Maps with mild filtering. In order to remove noise/obtain rather clean surfaces I have found that a few smooth mesh strength 1, a degree of decimate mesh and then a single smooth mesh strength 1 gives the best result that I have been able to obtain so far. The texture type is Diffuse map and the mapping mode is generic.
Nonetheless, the texture/image quality is somewhat lower compared to an orthomosaic created from the same images.
Is this to be expected?
Can anything be done to enhance the quality?
Does anyone have guidelines (beyond what the manual states) for when one should use normal map texture type and adaptive orthophoto mapping mode?
Are the any pros/cons for using a tiled model (other than the tiled format)?
Thanks