New Stereoscopic 3D Technique-Not Red/Blue


#21

Mike,
I wonder if there is a big difference between the types of glasses?

I am actually just holding lenses in front of my face and the depth effect is quite pronounced.

Now to be fair, I just started experimenting with this a couple of days ago and dont claim to understand it completely. Additionally, I think a higher end application would be more desireable for our use, whether that would be the polarizing glasses or perhaps the special monitors.

I have the animation in mp4 format, and wonder if that would make much difference over the flash version (fewer artifacts).

David


#22

I created it for red/cyan glasses, not Red/Blue

Cyan is a shade of blue (well blue-green really), it is probably only me that calls them red&blues.:slight_smile:

Not a bad render but I agree your offsets are a little funky. Like you are trying to give it too much depth. A couple of things I learned from the kind folks at Kerner Optical, backgrounds are super important to help define your depth, depending in the type of depth your trying to achieve shots generally work/look best when the whole subject is within the bounds of the screen, and do only what you need to get the desired effect (ie don’t go overboard).

IMO anaglyph is far less superior to any of the polarized approaches I have seen.

But like I said I am a total rookie at this too :slight_smile:


#23

There is a simple way to preview your results, and that’s crosseyed viewing. For those who who don’t know, what this is:
http://www.starosta.com/3dshowcase/ihelp.html
Create your virtual stereo camera, make a test render and put the right eye image to the left and the left eye image to the right. Cross your eyes and you’ll see if everything is OK.
Later on you can decide what kind of output you like.


#24

Are there other Plugins (purchased) out there besides the SVI version?

Thanks,
David

EDIT: here is the mp4 version MP4


#25

I made a script for this (though it doesn’t do polarized).
It was made for a video game my studio was doing in 3D, we used it for testing and researching what different values/settings were best (from within Max, instead of relying on programmers to adjust 3d settings in engine). It doesn’t really have much instruction and was made really for just a few people in mind; but you may find it useful.
http://www.thedour.net/scripts (“Render 3D v1.0”)
Click Download to the right of the title.

Once the script is launched you can click "Info" for some basic instructions.

All you need to do is have a camera in your scene, pick it, adjust 3d settings, configure your output.

Let me know you have questions!

#26

Andrew,

Do you know if there is an Anaglyph DirectX shader?

Just wondering…

Does your script take a long time to render too?
The SVI plugin sits and processes for the longest time before spitting out the frame.
I have sent them an email but they dont respond…
Thanks!

David


#27

David,

Not sure about a Directx shader but that is an interesting concept.  Many people take a single image, modify the color channels, and offset one of the images; however true Anaglyph should be 2 images captured from physically different positions (like our eyes) So I don't know that a shader would accomplish what you want even if there were one.

My script takes your current camera and actually creates 2 new cameras (offset using the various settings), the "right eye" and "left eye" images are each rendered and then the appropriate color channels/mixing happens and a composited image is kicked out. 

SO, it does render 2 times (once for each eye); depending on you initial scene render quality this could take awhile. (but no more than x2 of course)

I do provide a progress bar with text to let you know which "eye" image it is rendering, and also tell you when it is in the compositing phase.  The compositing time is very short, so I'd just worry about how long a single render is.

#28

So if the other options are chosen for rendering like side by side or interlaced, what technology is used to display them?

This is fascinating BTW…
David


#29

http://www.bigbug.com/stereo3d.htm

That link has the most straightforward examples.

When I was making the script we knew we'd be using anaglyph, but I imagine it may be pretty easy to add the additional features for side-by-side/interlaced.  I've not really read up on how interlaced is displayed to people (nor do I know the exact methods).

If you are considering moving the 3D direction for client work, you may consider the different anaglyph color types.  I'll share what I've experienced:
[b]red/cyan: [/b]Great 3D effect, causes more headaches, actual colors don't read well, reducing saturation can be helpful in red/cyan to reduce the "ghosting" (if you are having problems with that)
[b]green/magenta:[/b] Ok 3D, less headache, colors are virtually gone... might as well be black& white, generally not worth your time.
[b]blue/amber (aka [ColorCode](http://www.colorcode3d.com/3-D_Gallery.html)):[/b] Great 3D (though doesn't always pop quite like red/cyan), not a headache, excellent color representation.

I really liked blue/amber best.  Some other general tips:  Don't make the spacing of the "eyes" too far apart or it really messes with you.  Lower FOVs (below 60ish) seem to work better.

You may want to get pairs of 3D glasses for each color type, do some experiments, see what you like best.  :thumbsup:

EDIT: Added a “Stereo Pair” option to the script.


#30

Andrew,

When I use your script (thank you btw) the render times are about as long as with the SVI script, so at least that is consistent.

I can’t however get the same results with your script as with the SVI plugin.

Top Image created with SVI plugin.
Bottom image created with your script.

The red and cyan colors seem reversed (I have palayed with the settings and can get it to switch using the distance). I think the spacing need to go even smaller than .5" (at least in my applicatioin of it).

Let me know what you think…
David

SVI Image

Script Image


#31

Red/Cyan:
Well I looked into this, and it appears there are differing opinions on whether cyan/red goes first. Some tools (like the good ol’ Stereographer Max) give an option for whether cyan or red should come first. After looking at SVI and some other documents, the more common is cyan first: All this to say, my script had red first. I uploaded a NEW version to better fit with the commonly preferred method. (so your red/cyan reversal should fix itself)

    SVI vs other:
    I'd say SVI seems to be using a larger spacing (so yours may need to be larger even).  I downloaded SVI, they are doing a lot of different things.  I don't know what steps they are using.  Mine doesn't have any sort near/far planes; It appears they are changing some numbers/offsets based on the near/far you specify.
    
    I never claimed to be an expert but after looking at all the bells and whistles SVI has: I especially feel out of my league. :hmm:

#32

Andrew,

You know a LOT more than me, and I appreciate all your effort to help me out!
I will look at your revised script later today (if I can).
The SVI plugin has a lot of adjustments for sure.

I just wish they would respond to emails/questions.
it doesnt bolster my confidence too much…

Thanks,
David


#33

While looking into the differences between SVI and mine I stumbled across [color=Red]this amazing explanation[/color] by Louis Marcoux @ the Area.

I’m sure SVI is using many techniques talked about in these videos. I was using the “easy” method found in the first video… but not anymore :lightbulb

Render 3D v1.2 (yeah, updated again… sorry)
• Cameras now uses the “true” stereoscopic method.
• Has an “Auto Spacing” mode that uses the 1/30th distance “sweet-spot” rule. (this can be turned off and you can use custom values)
• Other various things to spice it up.


#34

Andrew,

It works much better…
There is an issue with ghosting now. I tried turning down the saturation to 75% but it was the same.
The depth effect is much improved.

Thanks,
David


#35

Looks like your spacing/offset is high, try moving your camera’s target to the middle of the skull (to set the convergence point there)


#36

Dave your offsets are crazy off :smiley: I think maybe try a different (lower) camera angle, too :wink:

@thedour I’ll have to check out that script your working on :slight_smile:


#37

Yeah JR, but I was trying to use his defaults first…

Thanks,
David


#38

Sorry all! :blush: I messed something up. When transitioning the script from “easy stereoscopy” to “true stereoscopy” I ended up still giving both render-time ‘eye’ cameras the same target (a carry over of “easy”). So they were both turned inward causing some crazy ghosting and unexpected results.

 new fixed version (100% verified... you are going to see the difference immediately)
 
 [[b]Render 3D v1.3[/b]](http://www.thedour.net/scripts)
 
 Thanks for all your persistence on this issue. 
 You guys are my favorite bug testers. :bowdown:

#39

Andrew,
I am the king of “bug finders”.
More than likely they are attracted to me…

Thanks,
David


#40

Big push in the news in the last couple of days regarding 3D TVs hitting the market. Took a glance at the Sony offering and they (I’m also guessing the others) are using frame sequential techniques with shutter glasses.

Guess the images are interlaced on the screens or are they progressive with a fast refresh? Guess the Nvidia setup would be the matching production setup for a workstation?

Thanks!