Compositing for distant aerial using depth map


I’m wondering if anyone has tips or a technique to do bird’s eye views (looking downward from the sky at about 45 deg) of large urban areas, like as much as the entire downtown area of a good size city (which might include something like 20x20 city blocks) such that maybe a lot of the repetitive things like street trees, can be done in post.

The main problem is that there are hundreds or thousands of trees over a great deal of perspective, and some blocked by buildings. I’ve seen many images like this, and they are probably rendered right into the scene, but I can’t image my humble computer not choking to death on this, even using super low-poly trees (they don’t need much detail since you are so far away). And I can’t imagine putting that many objects into a 3D scene.

So, I began to wonder if there is some solution out there that would use depth data to do some of the work for me. It seems possible that the grey scale value in the depth map could size the tree and would know when an object like a building closer to the camera will obscure the tree I’m putting in.

I dunno, seems pretty unlikely that exists. BTW, I’m using max 2009 and Vray typically.


can’t you just render a bunch of trees from the top view or even photographs and then just texture map them on spheres or cubes?


for reference:

Not a perfect example because they are showing a bit smaller area than I need to show, therefore I need even more trees.

What I’ve done in the past is to take a sphere and map it with something like the canopy of the tree, and with opacity maps and shadows it looks ok, but once you have a thousand trees or more, all casting shadows it get’s a little dicey (slow and sometimes crashy), even using vray proxies.

BUT, the point is to have them there in a convincing way, not to find a trick in photoshop, so if there’s a good way to do this I’m all ears. I should say that I’ve seen dozens of renders like what I’m going for, but I can’t figure out how, other than the slow way, and I can’t find a tutorial that shows this.

This is why I’m wondering about a method outside the renderer. The conundrum is that while the trees are absolutely essential to the image they’re not the subject, so committing all the processing time to them is a bummer.

I guess my wish is for something like scatter or fur but inside photoshop where I could “paint” the objects (especially if it could take a bunch of different tree objects and randomize them to eliminate repetition of the same exact thing). It seems that using a depth map from a render of the ground and the buildings, you would have enough info for a smart plugin to not only know what’s in front of what, but could size the objects being painted onto it to fake the perspective.


This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.