As mentioned in my post on the Severnaya Zemlya map i there for the first time used shape from shading techniques to improve a low quality elevation data set with the help of satellite imagery. Here a few more explanations on the matter.
The term shape from shading describes techniques that derive the geometry of an object from one or several shaded images of this object. Our human visual perception is to a large part based on shape from shading since stereo vision using both eyes to directly see the 3d form of an object only works at a very limited range and for far objects we rely on various indirect clues to assess the form of things we see, in particular shading. And due to many years of training our brain is extremely good at this. Still it is quite easy to confuse the brain in this matter and many optical illusions work this way.
For the computer shape from shading is difficult and in case of satellite images there are a number of aspects that further increase the difficulty:
- You usually only have a single image to work with.
- If you have multiple images they usually have the same or a very similar illumination due to the satellites’ sun synchronous orbits.
- If you have several images with different illumination you usually have other differences between the images (like seasonal changes) that mess with any interpretation of the shading.
As you can see this is really quite bad and therefore shape from shading is not usually considered feasible to create relief data from scratch. It is however possible to try improving an existing relief data set using shading information from satellite images. Read what needs to be considered to do this practically and how the results look. Below a quick look: on the left the original relief rendering and on the right the improved relief data: