GeoMipMapping - my approach

Started by
4 comments, last by martinperry 14 years, 2 months ago
I am currently working on geomipmapping. I want to solve cracks in a nice way. So I decided to use this: Image - LOD 4 Image - LOD 3 Edit: - added solution for corner Image - LOD corner solution As you can see, I solve possible cracks in pre-processing, when I generate indices. Positives: - all is preprocessed - LOD 1 can be connected to eg. LOD 10 :) - 1 chunk is drawn in one draw-call - while rendering, i dont need to solve neighbour LODs - I cant change LOD´s view distance very easily - easy to generate - just use recursion (only problem are corners, which must be done another way - I didnt solve them yet :)) Negatives: - require more triangles, if we connect LOD2 to LOD2 - little mess in my code :-D So far, I see, that there is more positives than negatives. I want to ask: is my approcach bad, or is acceptable ? [Edited by - martinperry on February 10, 2010 4:52:49 AM]
Advertisement
It's an interesting idea. My crack fixing patches up directly using a single polygon row, which can create artifacts in extreme cases. On the other hand, my experience is that more than 3 levels of LoD is totally pointless and so it doesn't much matter.
SlimDX | Ventspace Blog | Twitter | Diverse teams make better games. I am currently hiring capable C++ engine developers in Baltimore, MD.
Yep.. I use LOD 4 at max for very large terrain... with LOD 10 I only demonstrate possibility :)
I use a similar technique in my game with 3 level of details, I also suggest to take into account the terrain curvature in the boundaries to remove unnecessary vertices
Thodwros: Yep... I count with that as a third optimalization step :)
And Screen from almost - final app. (terrain is for now flat)

Image - Final - GeoMipMapping

This topic is closed to new replies.

Advertisement