CompuServe Thread

#rendering problem

7 messages in this thread
#111113From: Paul VickersJul 4, 1994 6:13 AM
3D Studio Rendering problem A corner of a room. A rectangualr wall object built with a box, with planar mapping Coordinates and a 24bit TGA texture of bricks, intersects in 60 degree angle with another wall of the same characteristics and a floor object built by getting a shape from the 2D shaper. A 30 mm camera runs parallel to the contiguous wall. When rendering this simple animation , on some few frames, the texture map of one of the walls appears distorted as if it had been pixilated or boxilated. It only occurs in one of the walls and only in some frames of the animation, as if the distortion was related to the angle in which the camer is targetting the wall. There does not seem to be a specific patter to the relation between the camera angle and the wall. This corner situation is repeated in several places within a model. Depending on the specific corner location we get different variations. In some corners there is no problem at all. In some corners the distortion occurs at the beginning of the turn, when the camera is perpendicular to the wall, sometimes it occurs in the frames when the camera is turning and the wall gets very close to the camera and sometimes it occurs at the end of the turn when the camer is parallel to the wall. Any explanation? we can provide example images if necessary. These problems have been noticed in version 2 as well as 3. Adrian Cook ajc@epic.co.uk
#111118From: ALAN IGLESIASJul 4, 1994 10:26 AM
Adrian, Do the faces you see this problem happening on by any chance share the same (or close to the same) space as a face (or faces) from another object? For example, do your walls intersect in such a way so that some faces may be (or are very close to being) co-planar? If this is the case, move the objects away from each other (at *least*) a couple of hundreths of a coordinate and render again, as the renderer needs to be able to discern exactly which face to render, and if they are too close together, it can get fooled. This is just a "shot in the dark", so to speak. I hope you find the problem. -Alan Iglesias
#111223From: Paul VickersJul 5, 1994 7:52 AM
Thanks Alan, I have just checked the object and it has no co-planar surfaces that are below about 10000 units apart from each other. This has raised the question about the scale of our objects. They are massive, this 10000 units is the width of a bounding wall so the bounding scene rectangle will be about 1000 times that. Any other ideas? Adrian Cook ajc@epic.co.uk
#111248From: Yost GroupJul 5, 1994 9:44 AM
If your objects are TOO large, then they'll overrun the floating point accuracy for the mapping coordinates. Try rescaling the scene to real-world coordinates, then reapply mapping coords and see what happens. – G
#111265From: ALAN IGLESIASJul 5, 1994 10:46 AM
Gary's ideas about scale, of course, are good things to try. I'm afraid there's not much more I can think of without seeing the file, though. Hope you figure this out, -Alan
#111122From: Yost GroupJul 4, 1994 11:06 AM
In addition to what Alan mentioned, there are a couple of dozen different "no-no's" that could be going on in your scene that might cause those kinds of artifacts. Most of them relate to coplanar geometry, overly large or overly small mapping coordinates, or a combination of the above. Without actually seeing the room model, I can't tell you what you're doing that's causing it. If you'd like to zip the file up with the maps and a text file explaining exactly what to look for, you could send it to me via binary email and I'll take a look at it for you when I get some free time this week. – G
#111129From: M. G. BATCHELORJul 4, 1994 11:42 AM
Is what you are describing situations where the walls (and thus mapping) are being distorted due to the focal length (w/a) and camera/object position ? If so, use a much higher res. map and/or stronger filtering, perhaps summed area mapping. It's hard to tell if this is what you are describing. BILL