Logo by lycium - Contribute your own Logo!

END OF AN ERA, FRACTALFORUMS.COM IS CONTINUED ON FRACTALFORUMS.ORG

it was a great time but no longer maintainable by c.Kleinhuis contact him for any data retrieval,
thanks and see you perhaps in 10 years again

this forum will stay online for reference
News: Check out the originating "3d Mandelbulb" thread here
 
*
Welcome, Guest. Please login or register. March 28, 2024, 08:57:08 PM


Login with username, password and session length


The All New FractalForums is now in Public Beta Testing! Visit FractalForums.org and check it out!


Pages: [1]   Go Down
  Print  
Share this topic on DiggShare this topic on FacebookShare this topic on GoogleShare this topic on RedditShare this topic on StumbleUponShare this topic on Twitter
Author Topic: Stereo settings  (Read 3367 times)
Description: correlation to S3D film language
0 Members and 1 Guest are viewing this topic.
bengvc
Alien
***
Posts: 34


« on: February 13, 2012, 03:24:18 AM »

I learned stereo from the live-action realm, so our language may differ a bit, but trying to dial in my stereo template to avoid (minimize) edge violations and control overall stereo budget. How do interaxial and convergence relate to the three variables, screen distance, image width, and minimal distance? I'm playing back on a 50" active 3D plasma, sitting about 2 meters away, so the default for screen distance can stay put, correct? Presently, I can render L/R images, bring them into After Effects, apply 3D Glasses for testing stereo in anaglyph or out to interlaced 3D Zalman, but sure would be nice to do a quick test render for stereo template within Mandelbulb 3D. Thanks for direction. Amazing app, finally able to explore.
Logged
Jesse
Download Section
Fractal Schemer
*
Posts: 1013


« Reply #1 on: February 13, 2012, 09:14:11 PM »

Yep, seems that it fits the default values in stereo mode.
If the overlap is to much in nearby objects, you could increase the 'Minimal distance' value that defines how near (also in meters) the closest object appears in front of you.

Is the usage known? ..render the image as usual, this would be the "right eye" image. (save it)
Then click on "Render left eye image" and you should get the image for the left eye.  Save it too, combining them somehow is not up to me  wink
In animation rendering, just check the "Render stereo animation" option and both images will be calculated and saved automatically.
« Last Edit: February 13, 2012, 09:42:13 PM by Jesse » Logged
marius
Fractal Lover
**
Posts: 206


« Reply #2 on: February 14, 2012, 02:09:09 AM »

Yep, seems that it fits the default values in stereo mode.
If the overlap is to much in nearby objects, you could increase the 'Minimal distance' value that defines how near (also in meters) the closest object appears in front of you.

Is the usage known? ..render the image as usual, this would be the "right eye" image. (save it)
Then click on "Render left eye image" and you should get the image for the left eye.  Save it too, combining them somehow is not up to me  wink
In animation rendering, just check the "Render stereo animation" option and both images will be calculated and saved automatically.

Jesse, do you do a simple side-step for the left eye (symmetric frustums) or keep the projection plane exactly the same (asymmetric)? You'd want something like the diagram below and then parameterize the eye separation and the ratio thereof to the projection plane.

Logged
bengvc
Alien
***
Posts: 34


« Reply #3 on: February 14, 2012, 03:12:19 AM »

In Marius' diagram the cameras are converged on the screen plane, which stated more accurately, they're converged on distance to the nearest object, placing them at screen plane.

If I'm getting it, in my example of anticipated venue being on my 1.5 meter wide active display, and I'm sitting 2 meters away, then if I set nearest distance also to 2 meters, nothing will forward project, i.e. window mode. Correct? And if I find the overall front and rear parallax too aggressive, I either increase the screen width or decrease distance from screen?

I suppose because S3D live action deals with real distances (to farthest and nearest object, between cameras), and because measurement of CG space is arbitrary, these parameters for screen width and distance from screen provide a kind of safety net. I'm admittedly confused how you can pin distance to nearest object, or is convergence constantly being callibrated against the relationship between camera and nearest object? That would appear to invite highly bothersome 3D, which I'm not seeing in clips out there. I hope the question is clear; if I fly the camera directly toward a protruding spire, how does "nearest distance" deal with the increasingly forward projected subject matter without inviting major divergence problems in the background?
Logged
Jesse
Download Section
Fractal Schemer
*
Posts: 1013


« Reply #4 on: February 14, 2012, 11:23:33 AM »

Jesse, do you do a simple side-step for the left eye (symmetric frustums) or keep the projection plane exactly the same (asymmetric)?
 You'd want something like the diagram below and then parameterize the eye separation and the ratio thereof to the projection plane.

Yes, it is done similar to the graph in the image, but the viewvec is rotated instead of direct changing the FOV offsets for the projection plane.
It is indeed a bit more complicated because the camera uses a starting plane that would not result in a camera-point without the rectangular camera setting, and with rectangular the rotation does not represent a correct FOV modification.. and there is another issue.
So by now it is a compromise, but for common usage it should be ok.

If I'm getting it, in my example of anticipated venue being on my 1.5 meter wide active display, and I'm sitting 2 meters away, then if I set nearest distance also to 2 meters, nothing will forward project, i.e. window mode. Correct? And if I find the overall front and rear parallax too aggressive, I either increase the screen width or decrease distance from screen?

Screen width and distance should not be changed, use the 'Minimal distance' instead.
Add: Ok, if you want to set infinity more towards the screen to minimize offsets, you can increase screen distance and size by multiplying both with the same number.

Quote
I'm admittedly confused how you can pin distance to nearest object, or is convergence constantly being callibrated against the relationship between camera and nearest object?

My explanation was a bit misleading, i meant that the minimal distance parameter defines the nearest point an object can appear (that would be start of the rendering ray at a certain distance from the cameras point).
There is no automatic calibration of the nearest object in the image.
« Last Edit: February 14, 2012, 03:47:48 PM by Jesse » Logged
bengvc
Alien
***
Posts: 34


« Reply #5 on: February 15, 2012, 05:04:13 AM »

I ran some interesting tests in Mandelbulb to get a handle on how the stereo parameters relate, learned much and discovered some issues. If the 3D works, some will argue, why fix it? The brain can and does deal with a good amount of bad 3D in the real world, but in stereoscopics geometric misalignments between left and right eyes detract from the quality, at best, and can cause headaches at worst. So, anybody here interested in overall quality in their fractal masterpieces intended for stereoscopic output, it all counts.

Testing workflow: I rendered seven pairs of stereo images, bumping up each parameter individually, then sometimes in pairs, according to a logic that tested for positive and negative parallax percentages, as well as overall stereo budget, incrementing the aggressiveness of the stereo effect. This can be seen (subjectively) by viewing in stereo, but is better analyzed with glasses off and by counting pixels or calculating screen width percentages. Each left eye file was dropped into a left eye comp in After Effects, same for the right eye. To quickly see which stereo pairs I was viewing in comparisons to come, I added text layers to each image indicating Screen Distance, Screen Width, and Min. Distance settings, did this for the right eye as well, not to stab one eye while viewing in 3D. I then duplicated both L and R eye comps, this to provide two L/R comps in a master comp used to quickly switch between the different setups. With all seven stereo pairs in each of the four comps, two left eye and two right eye, I could set visibility to the respective layers in each of the seven stereo pairs, making it easy to cover all combinations comparing 1 to 2, 2 to 3, 1 to 3, and so on. I'm able to make sense of some things, less of others.

As min. distance values approach screen distance, forward project subject matter is pushed back to screen plane. I specifically used a big pointy tricom fractal pointing straight to camera to provide something easy to fix on and that allowed major forward projection without clipping the camera frustum. It took a bit of squirming around before the Zalman to even begin to accomodate the first stereo pair, using the default settings of 2 screen distance, 1 screen width, and .5 min. distance. Forward projecting any subject matter 50%, even a pointy thing in the middle of the screen, is asking for headaches. Making your eyes converge that much to accomodate the left and right images is one thing, but then the overall stereo budget in the scene can make things even harder to take in. If you push the subject matter back toward the screen, still poking out by say 25% (min. distance is now 3/4 the screen distance value), then if the rear projected subject matter diverges too much, you're still inviting headaches. I work with a lot of 3D, so my eye muscles can take more than many (2-3% of the population can't see 3D at all, and about 15% are "stereo challenged"). Depending on the size of objects, my eyes can accomodate fairly aggressive 3D, but the default stereo settings yielded so much parallax, I had a hard time getting the entire Tricom to come together. You don't want the viewer to have to chose which part of an object or a scene to accomodate, the entire scene wants to be accommodated. This is how viewing 3D differs from real world experience, where we naturally target subject matter, blocking out everything else.

Increasing the screen width, in combination with increasing the min. distance, began to tame the 3D to acceptability (to my eyes). Oddly, setting of 1,2,1 for screen distance, screen width, and min. distance respectively yielded identical results to 2,2,2, save some minor artifacting. Would have to test that again to see if this condition, screen width exceeding screen distance, with both values going up or down together renders similar results.

A final observation, I noticed not a small amount of vertical misalignment between left and right views. The brain can deal with some of that, but it's not something you want, our eyes are side by side for a reason. Corner pinning can correct for that, for those that care. I understand this forum is about fractals, not stereoscopics, but 3D, that is S3D is a powerful aspect of exploring the Mandelbulb, it's worth getting right. At least now I've got somewhat of a bearing as I head into the forest of formulas.
Logged
marius
Fractal Lover
**
Posts: 206


« Reply #6 on: February 15, 2012, 05:55:38 AM »

...

fwiw, a while back I put an interlaced stereo output option in http://code.google.com/p/boxplorer2, since I also happen to have a Zalman. Depending on how good your videocard(s) are, you can fly around the simpler formulae in realtime real3d  alien
Logged
bengvc
Alien
***
Posts: 34


« Reply #7 on: February 15, 2012, 06:00:54 AM »

Nice, just what the doctor ordered, plenty fast video card. Thanks.
Logged
Jesse
Download Section
Fractal Schemer
*
Posts: 1013


« Reply #8 on: February 15, 2012, 06:21:13 PM »

I have rewritten the stereo transformation because the offset between the "viewplane" (sarting plane of raymarching) and the camera position was not taken into account.
The difference is not very big, depends on the settings, but i hope some issues will be fixed.

Dunno why there should be a tilt on vertical lines, it is normal if the lines are pointing slightly towards you that you get a tilted line on the left-eye image, even the right-eye one shows no tilt... or you need to show examples.
Logged
bengvc
Alien
***
Posts: 34


« Reply #9 on: February 15, 2012, 09:14:30 PM »

Jesse, I've worked up a layered Photoshop file, each layer containing 1 of 7 stereopairs displayed in overlay. The titles of the layers are 3 numbers reflecting the settings for screen distance, image width, and min. distance. Rules align on features to accentuate vertical disparity issue, toggling visibility of layers reveals changes. I don't want to take up space attaching a 12mb psd file, can send you a download link, but we should stay full res in Photoshop to adequately discern 3D. You emailed me before offline, but I don't see your email, gone, do you still have mine?, I'll send the link.

Thanks for caring about the stereo!
Logged
Jesse
Download Section
Fractal Schemer
*
Posts: 1013


« Reply #10 on: February 15, 2012, 10:44:43 PM »

Hmm, have not got any new private message?

Anyway, you can test the updated version ( 1.7.8 ) if the behaviour is better now.
« Last Edit: February 15, 2012, 10:46:38 PM by Jesse » Logged
bengvc
Alien
***
Posts: 34


« Reply #11 on: February 15, 2012, 10:51:07 PM »

My tests, now on the old build, are arguably moot, but it might be revealing nonetheless, only if you have the time. The file turns out is more like 7MB. Let me know how I can get it to you.

I look forward to testing the latest and hopefully greatest. Thanks.
Logged
Jesse
Download Section
Fractal Schemer
*
Posts: 1013


« Reply #12 on: February 15, 2012, 11:23:13 PM »

I am sorry that you did so much testing on that old version, i knew that there was a kind of bug as i told, but had to much other things to do that i fixed it only yet, while this topic is disussed again and things are clearer.
I can only hope that this is fixed now, always thanks if someone has the time for more intensive testings and bugreporting!
Logged
Pages: [1]   Go Down
  Print  
 
Jump to:  

Related Topics
Subject Started by Replies Views Last post
saving stereo settings Mandelbulb 3d bengvc 6 1867 Last post February 26, 2012, 09:37:55 AM
by bib
Anyone else noticed problems with Advanced Settings dialog box? Fractal eXtreme simon.snake 1 4695 Last post May 30, 2013, 10:56:34 PM
by kallisti
Keeping settings across projects Mandelbulb 3d rurik2000 0 1428 Last post February 24, 2014, 02:26:32 AM
by rurik2000
stereo settings and 360 panorama Mandelbulb 3d barcud 1 3699 Last post October 25, 2014, 02:23:36 AM
by NickVenden
stereo settings question mndelbulb 1,9,8 Mandelbulb 3d Frank Krischick 1 1326 Last post January 04, 2015, 09:22:00 AM
by Rathinagiri

Powered by MySQL Powered by PHP Powered by SMF 1.1.21 | SMF © 2015, Simple Machines

Valid XHTML 1.0! Valid CSS! Dilber MC Theme by HarzeM
Page created in 0.241 seconds with 25 queries. (Pretty URLs adds 0.011s, 2q)