You'll need red and green glasses to view this, but maybe like me you never throw away a pair...
(http://api.ning.com/files/uLAN9BGRX*vxMunhQ7SmkoTAwPFL7Rp9gWcRKkZ8G11UOWt-Y4Jd1tBpaC56G4K3t5OBq3WUzQJKiBMe8Tsa8uwAr1wXoanL/pluto3D2.jpg?width=737&height=552)
nice!
you may want to bring the reflection of the sun down a bit - it draws the eye more than you'd want it to.
It's cool, but you have broken the first rule of stereoscopy - never break the frame with something in front of the convergence point. It messes with the brain too much.
Your foreground rocks are closer than the "frame", which causes our brains to get confused because on the left and right edges, each eye sees a point that does not exist when seen with the other eye.
The advice is all good. The picture was just a test of concept, though to see if it would even work at all. Turns out to be dead easy.
The image, by the way, appears just dandy when I look at with my glasses...but maybe that's just me.
It looks pretty good. I didn't notice the foreground rocks breaking the frame, but my eyes are going bad. Is the sun set as the focus point and did you just move one eye (camera) a small distance to the side?
Quote from: mcmiller on December 18, 2010, 12:57:23 PM
It looks pretty good. I didn't notice the foreground rocks breaking the frame, but my eyes are going bad. Is the sun set as the focus point and did you just move one eye (camera) a small distance to the side?
Frankly, I don't think that anything in the landscape even
could break the frame!
I created one rendering and then moved the camera position over the smallest distance I could (human eyes are only about 2.5 in apart) and then did a second rendering. The small, free program Anamaker then combined the two into a 3D anaglyph.
What njen is saying is this: your foreground appears to be nearer than the screen plane. Objects nearer than the screen plane are OK as long as they don't touch the sides of the image, but when they touch the sides they appear to be cut off in different places from one eye to the other, and this can be jarring. If everything is further than the screen plane you don't have this problem and then the screen appears like a window into the world behind it.
To push the landscape behind the screen plane you could adjust the rotation of each camera so that they point very slightly together (leaving the positions how you already have them). You would want to point them inwards just enough so that the foreground never quite reaches convergence. Another method that allows more control post-render is to render each image slightly wider so that you can slide each image afterwards in your 2D program and then crop inwards. Actually I'd probably go with that second method every time.
Having said that, red-green or red-blue anaglyphs probably make everything I've just said less important. Because of the ghosting problems and general weird look due to the different hues I can see why you want to put the horizon at the convergence plane (screen plane), because you want the horizon to be the easiest to focus on. So maybe I wouldn't change a thing after all...
Quote from: Matt on December 18, 2010, 06:30:52 PM
What njen is saying is this: your foreground appears to be nearer than the screen plane. Objects nearer than the screen plane are OK as long as they don't touch the sides of the image, but when they touch the sides they appear to be cut off in different places from one eye to the other, and this can be jarring. If everything is further than the screen plane you don't have this problem and then the screen appears like a window into the world behind it.
I realize what he was saying. The problem really lies in the fact that unless you have a landscape in which all foreground features are crowded within the angle of sight or just far enough away (so nothing comes near the edges of the picture plane) you will always have this sort of problem with some features. But then you'd have a pretty artificial-looking landscape, too. (This whole problem exists largely because the frame itself is an artifact that doesn't exist when you are looking a scene with the naked eye.)
QuoteTo push the landscape behind the screen plane you could adjust the rotation of each camera so that they point very slightly together (leaving the positions how you already have them). You would want to point them inwards just enough so that the foreground never quite reaches convergence.
In other words, look at it more cross-eyed! ::)
QuoteHaving said that, red-green or red-blue anaglyphs probably make everything I've just said less important. Because of the ghosting problems and general weird look due to the different hues I can see why you want to put the horizon at the convergence plane (screen plane), because you want the horizon to be the easiest to focus on. So maybe I wouldn't change a thing after all...
I hadn't seen anyone try a 3D image with TG before so I thought I'd give it a shot just to see if it would even work. This first attempt certainly does have some technical problems (for instance, the two cameras are actually several feet apart, not a few inches). The next time I have a few minutes (ha!) to mess around, I'll try some of your suggestions. Might be interesting if someone had the time and technical adroitness (which I have precious little of on both counts, God knows) to try a 3D animation. Of course, I'd not be remotely surprised to learn that someone in here has already done that.
PS---
I'm glad to see that I'm evidently not the only one who has never thrown away a pair of old 3D glasses! ;D
Quote from: blackcat on December 19, 2010, 08:19:26 AM
I realize what he was saying. The problem really lies in the fact that unless you have a landscape in which all foreground features are crowded within the angle of sight or just far enough away (so nothing comes near the edges of the picture plane) you will always have this sort of problem with some features.
I am not sure you do know what I am saying. As someone who has worked on stereoscopic feature films, you have to first understand about the concept of what is known as the convergence point. It is the point at which everything in front of the point is in front of the frame (edge of the image), and everything behind the convergence point is behind the frame. Not breaking the frame with objects in front of the convergence point is the rule that what 99% of stereoscopic feature films
follow.
You can fill your image with a landscape, but simply do not have any of it in front of the convergence point.
Quote from: blackcat on December 19, 2010, 08:19:26 AM
In other words, look at it more cross-eyed! ::)
Again, I am not sure you understand about stereoscopy properly. That rotating the cameras do not make the image "cross eyed", but rather shift the convergence point forward or backward.
Here is a really rough example. In the first image, the convergence point is the furthest object away from the camera (the left torus knot). Every other object is in front of the convergence point and safely not touching the frame except for the checkered teapot, which looks strange to the eyes in how part of it is being cut off.
[attachimg=1]
In the second image, the convergence point is the small sphere in the middle. Now all of the big objects that are touching the frame are behind the convergence point, and look totally natural, while the small teapot on the left is the only object in front of the convergence point, but is not touching the frame so it still looks good.
[attachimg=2]
By the way, the trendy way to produce stereo images now days is to render the cameras parallel to each other, and then adjust the convergence point in post software afterwards (Nuke for example). That was how I did these images. Both of them are using the same 2 source images, I only adjusted the offset.
One easy way you can assess this rule (assuming left is red, and right is blue), is if you see an object that is red on the right edge, and blue on the left, it shouldn't be touching the frame.
Quote from: njen on December 19, 2010, 03:33:04 PM
Quote from: blackcat on December 19, 2010, 08:19:26 AM
I realize what he was saying. The problem really lies in the fact that unless you have a landscape in which all foreground features are crowded within the angle of sight or just far enough away (so nothing comes near the edges of the picture plane) you will always have this sort of problem with some features.
I am not sure you do know what I am saying. As someone who has worked on stereoscopic feature films, you have to first understand about the concept of what is known as the convergence point. It is the point at which everything in front of the point is in front of the frame (edge of the image), and everything behind the convergence point is behind the frame. Not breaking the frame with objects in front of the convergence point is the rule that what 99% of stereoscopic feature films
follow.
I see what you're saying now! (The terminology and some of the concept threw me at first since both are little different than similar terms and usages in ordinary perspective drawing...which is based on the assumption of a single, i.e., one-eyed, viewpoint...or at least just different enough to confuse me---an easy enough process in even the best of circumstances. :( )
QuoteYou can fill your image with a landscape, but simply do not have any of it in front of the convergence point.
Quote from: blackcat on December 19, 2010, 08:19:26 AM
In other words, look at it more cross-eyed! ::)
Again, I am not sure you understand about stereoscopy properly. That rotating the cameras do not make the image "cross eyed", but rather shift the convergence point forward or backward.
The
image isn't cross-eyed, of course. I just was trying to describe the equivalent effect of pivoting the cameras toward or away (wall-eyed in that case) from one another. It's what you do when you are trying to merge two dots together in a stereoscopic image or in trying to make a stereogram work.
Anyway, thanks for the explanation and examples! As I said, my experience in ordinary geometric perspective (which, as I suggested, is based on a unique viewpoint) got in the way. (I can---with just a straight edge and a pencil---draw a room with vaulted arches and a spiral staircase, all reflected in a mirror tipped at an angle against a wall, with the proper shadows cast by a light hanging from the ceiling. But the geometries involved in stereoscopy are new to me and I got myself flummoxed by both the terminology and concepts. :o ) Your descriptions are very much appreciated! I'd like to see some TG users try their own 3D renderings---it would be very interesting to see what more successful experiments than mine might look like!
Nice image , Blackcat :)
This image from Hannes came to my mind:
http://forums.planetside.co.uk/index.php?topic=9578.0
And Njen thanks for your examples i learned something too :)
Quote from: Kadri on December 20, 2010, 01:50:50 PM
Nice image , Blackcat :)
This image from Hannes came to my mind:
http://forums.planetside.co.uk/index.php?topic=9578.0
And Njen thanks for your examples i learned something too :)
Very nice! :)
I wish I had time to play more with this sort of thing!
Quote from: blackcat on December 20, 2010, 07:30:13 AM
I'd like to see some TG users try their own 3D renderings---it would be very interesting to see what more successful experiments than mine might look like!
Go and see Tron Legacy in 3D :)
I am going to miss a lot a cool stuff. I can't see in 3D.
Yes sports are very difficult. I have no way to see how far away things are.
Quote from: njeneb on December 21, 2010, 08:23:12 AM
I am going to miss a lot a cool stuff. I can't see in 3D.
Yes sports are very difficult. I have no way to see how far away things are.
Andre de Toth, the director who made one of the first successful 3D movies, "House of Wax", couldn't see in 3D, either!
I think it helps me working with 3D modeling software.