To make real-time terrain rendering engine, assuming that
I can implement by myself the algorithm (geomipmapping,clipmapping or topovista …not decided yet) and using GLSL - is there any advantage using OpenSceneGraph (or other) versus doing it myself?
If you don’t have many objects on terrain ,then using dedicated terrain rendering algorithm will be better.
But if your terrain will introduce vertical cliffs, caves, lots of buildings and trees, then OSG will most likely work better.
It all depends upon what your goals are, and how much time you have on your hands.
It takes years to write a fully functional and scalable database generation tool and rendering system. Most of the published papers/algorithms don’t scratch the surface when it comes down to making things really scale up, and handling of mosaics of geospatial imagery and dems. The VirtualPlanetBuilder/ OpenSceneGraph combination solves both the building and rendering and comfortable scales up to handling terabytes of data.
However, if you just want to render a chunk of terrain, a couple k square then sure it’s pretty straight forward to go implement your own algorithm. However, given that there are off-the-shelf solutions out there already you aren’t gaining anything technical, so you really have to justify it in terms of about learning about terrain rendering and coding rather than actually solving the problem in hand.
You could of course still using something like the OpenSceneGraph and implement your own terrain algorithm on top. There are plenty of reasons to use a high level toolkit beyond extras like terrain rendering that they might provide.