![]() ![]() So a pronounced improvement could be made here by feeding more relief information to the procedural texturing techniques (and of course by better landcover resolution in the first place). Our current way of texturing terrain reflects this very poorly - if it's not in the vector data, we can't do it, even if we have the relief. But in addition note how reality shows many structures aligning with the relief of the terrain - terraced fields in the marked region for instance following the equal height lines of the terrain, trees pushing out tendrils with the valleys. Agriculture and village do have knife-sharp transitions again, and no smoothing or transition effect would in fact fix the problem here. Here we have a data problem - the landcover data doesn't really reflect what's really there. All this could be done easily if we were able to target regional textures better.Ī rather less charitable comparison can be made in Hawaii, here Maui for real In fact, asked as to what would need to be done to make FG look more like the actual scene, my answer would not involve blending vector seams at all - it would involve mainly balancing colors better (especially between village and surrounding agriculture), and then adding trees to the agricultural sector. However, in most cases we could actually mimic those, by populating the service roads shown on the agriculture textures with trees for instance, or, in the case of the uppermost circle, by using a more detailed cropwood texture. Problematic areas are indicated with circles - reality has structures which are smaller than the resolution of the CORINE data. CORINE does in fact very well representing this kind of terrain.Ĭonsider for illustration a direct comparison for an even more challenging case - Grenoble transiting into the foothills of the Alps, first for real:Īgain, the arrows indicate knife-sharp transitions which occur in both cases. Note the transitions between forest and agriculture, or between village and forest, or between village and agriculture indicated by the arrows - they're all knife-sharp lines which would be perfectly well represented by the vector landclass data. ![]() Here's an image rather characteristic of central Europe (northern Germany in fact, all aerial images from google earth): I'm going to argue that this is largely not the case. Getting close to the terrain, it can be made to look almost arbitrary good, but there are issues with the resolution of the vector data when looking from farther away.Ī wide-spread perception is that our largest problem are the vector seams of the different landclasses, leading to knife-sharp transitions between the different landclasses, and that blurring them, blending one landclass into the other, would improve the situation. Which doesn't mean we couldn't do better.Ĭurrently we use landclass vector data where individual landclasses get assigned textures or procedural texture definitions, and these can vary region by region. in addition to widespread use of CORINE data for the Scenery 2.0, so the situation has in fact improved quite dramatically in the last years. Since my last writeup we've acquired a whole bunch of very useful techniques - regional texturing, procedural texturing, object placement masks, pixel postprocessing effects. I'd like to present a few thoughts what I think is the current holdup which prevents FG terrain from looking better than it now does. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |