3D Anaglyphs (Red/Blue Glasses)

DanaTA

Horo said:

I'm a great fan of anaglyphs, I started by photographing a scene a bit left and a bit right, made them red and cyan monochrome and assembled them in Photoshop, even cylindrical and spherical HDRI panoramas (including QTVRs) - though spherical are a bit critical because it swaps at the zenith and nadir, but cylindrical are fine. I also rendered some anaglyphs early on by combining the renders. It turned out that HDRShop gives the better result than Photoshop, even for normal renders (not HDRI).

Meanwhile, we have the anaglyph lens system for Bryce and an anaglyph can be rendered in one go, the "eye"-distance can be adjusted so it's relatively easy to exaggerate or bring the objects out of the display towards the beholder. I have a few in the Daz gallery: http://www.daz3d.com/gallery/#galleries/1987 and some on my website (Raytracing > Gallery 1 (and 2) > Anaglyphs).

I love those!  Very impressive.  My favorites are the Knotted Strip and the Pimpled Band.  The Bryce mountain scened are very good, too.  The one at the top of the page is great.  I'd love to see it fill my screen, though!

Dana

barbult

algovincian said:

barbult said:

Here is one I did today. I use StereoPhoto Maker to create the anaglyph image.

Edit: Oh wow, the forum image compression really adds bad artifacts to the anaglyph JPEG image I attached. I deleted it and attached a PNG file instead.

That looks great, Barbult! Out of curiosity, did you leave your cameras parallel, or toe them in? I know that theoretically they are supposed to be parallel, but in the image I posted I toed them in because I wanted the chic to be solidly in focus. Theoretically, most people will say that toed in will give you zero parallax depth, but it definitely looks 3D to me (and is in good focus/easy on my eyes).

- Greg

My cameras were parallel.

Taoz

Murgatroyd said:

Taozen said:

Maybe some day 3D programs will be able to create renders in stereogram format, like this:

(don't look at the picture, but "behind" it as if you're looking at something far away)

If you can render your scene as a depth map, there are any number of programs out there that can turn it into a stereogram for you.

Will take a look at that. Thanks!

kaotkbliss

I've read that these Pro ANA glasses are the best to use on Acer monitors for the least amount of ghosting (seeing double images while looking through the glasses) I did purchase a couple pairs of them for using nVidia's anaglyphic 3d gaming and they do work great, and they are dirt cheap :)

https://www.amazon.com/gp/product/B0036NP3CS/ref=oh_aui_detailpage_o06_s00?ie=UTF8&psc=1

Horo

 Go to my wbsite, Raytracing > Gallery 2 > Anaglyphs (for the Knotted Strips Anaglyph there's even an animation). 

DanaTA said:

Horo said:

I love those!  Very impressive.  My favorites are the Knotted Strip and the Pimpled Band.  The Bryce mountain scened are very good, too.  The one at the top of the page is great.  I'd love to see it fill my screen, though!

Dana

 

DanaTA

Horo said:

 Go to my wbsite, Raytracing > Gallery 2 > Anaglyphs (for the Knotted Strips Anaglyph there's even an animation). 

DanaTA said:

Horo said:

I love those!  Very impressive.  My favorites are the Knotted Strip and the Pimpled Band.  The Bryce mountain scened are very good, too.  The one at the top of the page is great.  I'd love to see it fill my screen, though!

Dana

 

Cool, thanks!  I bookmarked your site.

Dana

barbult

Here is another one. Cameras are parallel again.

 

algovincian

I have trouble (more ghosting) getting images like this one to focus when the cameras are parallel. Maybe I'll render out a direct comparison as an experiment when I get a chance.

- Greg

Horo

DanaTA - you're welcome.

barbult - one thing I find very important is that whatever meets the image frame must be kept at or behind the screen. Here, everything is fine except the front of the bike. I'm not so sure how much parallel or toe-in influence the perception. Usually, a bit of toe-in is more natural. On the other hand, the 3D effect is the fun, not how natural it looks.

wiz

CypherFOX said:

I would love to see an animation, though...like the infinitely odd image above just...rotating in place.  It'd be fascinating to see how that looked, in 'Magic Eye' format.  but with renders...)

You can't really have a render in Magic Eye format. The process doesn't lend itslef to shaded imagery. THink "depth maps".

wiz

algovincian said:

I know there's a ton of software out there to do this, but in this case, I just used Photoshop (CS2) to create the final image from two different views of the scene rendered using my NPR algorithms.

You really want to use some of that software, like SterepPhoto Maker. It has anaglyph ghost reduction.

algovincian

wiz said:

algovincian said:

I know there's a ton of software out there to do this, but in this case, I just used Photoshop (CS2) to create the final image from two different views of the scene rendered using my NPR algorithms.

You really want to use some of that software, like SterepPhoto Maker. It has anaglyph ghost reduction.

To be clear, wiz, I see limited ghosting in the image I posted to start the thread. I see a lot of ghosting and have trouble getting barbult's image to focus.

What do you (and others) see? Are my eyes f*!$&d up? Oh noes . . . I think I may be toed-in! lol

- Greg

 

barbult

Horo said:

barbult - one thing I find very important is that whatever meets the image frame must be kept at or behind the screen. Here, everything is fine except the front of the bike. I'm not so sure how much parallel or toe-in influence the perception. Usually, a bit of toe-in is more natural. On the other hand, the 3D effect is the fun, not how natural it looks.

Horo, I'm not sure what "whatever meets the image frame" means, or why you feel it must be behind the screen. Is it because you feel that it is easier to focus on when it recedes into the screen?

I can see my image jumping out from the screen just fine. The entire bike jumps out from the screen. StereoPhoto Maker has controls to adjust the separation to control what part of the image recedes and what part jumps out. I chose to leave this one at the automatic setting. My husband has a 3D camera and takes thousands of 3D photos. I'm very practiced at looking at 3D images (usually in full color on the 3D TV, not anaglyph). I wonder if that makes a difference in my ease of focusing on the popping out images.

barbult

algovincian said:

wiz said:

algovincian said:

I know there's a ton of software out there to do this, but in this case, I just used Photoshop (CS2) to create the final image from two different views of the scene rendered using my NPR algorithms.

You really want to use some of that software, like SterepPhoto Maker. It has anaglyph ghost reduction.

To be clear, wiz, I see limited ghosting in the image I posted to start the thread. I see a lot of ghosting and have trouble getting barbult's image to focus.

What do you (and others) see? Are my eyes f*!$&d up? Oh noes . . . I think I may be toed-in! lol

- Greg

 

Greg, one difference between your image and mine is that yours is a grayscale image converted to anaglyph, I believe. Mine is full color converted to anaglyph. The color images are much harder to deal with, because any red and blue in the original image cause ghosting issues. Also, mine jumps out from the screen a lot and has more separation between left and right eye than most of your image. The color calibration of your monitor and the glasses that you use may also lead to ghosting, if they do not match well. I'm going to convert mine to grayscale and adjust the plane of focus with StereoPhoto Maker and see if that makes it easier for your to see.

Edited to add:

I also think I used too much eye separation on my cameras, making it harder to focus on. I'm going to redo it again and see if it looks better to you. I'll post another new one in a few minutes.

Edited again: Nope, I think my camera separation is pretty good. I was confusing myself with inches and centimeters there for a second. I have 60mm separation, which is actually a little low for average interocular distance. So, I'm going to leave it alone for now.

 

Horo

What I mean is if any object that comes out of the screen towards the beholder, this object should not touch the picture frame or border because the picture border is level with the screen surface and an object that protrudes and is cut off by the border creates an "impossibility" and that part gets difficult to perceive correctly as 3D. In the scooter image, as long as you don't concentrate on the lower edge of the image where the protruding scooter hits the border, everything protruding looks just great: head and arms but the leg already irritates - at least me. Strangely enough, an object that receeds into the screen is cut by any of the image borders does not create this feeling.

 

barbult

Horo said:

What I mean is if any object that comes out of the screen towards the beholder, this object should not touch the picture frame or border because the picture border is level with the screen surface and an object that protrudes and is cut off by the border creates an "impossibility" and that part gets difficult to perceive correctly as 3D. In the scooter image, as long as you don't concentrate on the lower edge of the image where the protruding scooter hits the border, everything protruding looks just great: head and arms but the leg already irritates - at least me. Strangely enough, an object that receeds into the screen is cut by any of the image borders does not create this feeling.

Thanks for the clarification, Horo. Yes, I do see what you mean. I think my brain tends to ignore that and focus on the center of the image. I'll try to be more aware of that issue in the future.

algovincian

barbult said:

Greg, one difference between your image and mine is that yours is a grayscale image converted to anaglyph, I believe. Mine is full color converted to anaglyph. The color images are much harder to deal with, because any red and blue in the original image cause ghosting issues. Also, mine jumps out from the screen a lot and has more separation between left and right eye than most of your image. The color calibration of your monitor and the glasses that you use may also lead to ghosting, if they do not match well. I'm going to convert mine to grayscale and adjust the plane of focus with StereoPhoto Maker and see if that makes it easier for your to see.

Definitely clears up the ghosting I was seeing substantially. It looks like the background objects are shifted now majorly now, rather than the bike/rider as in the last image, too. Thanks for taking the time and uploading, barbult. Obviously, grayscale is simpler and better suited for red/blue anaglyphs (which is why I thought some of my sketches might make for good content).

I've downloaded StereoPhoto Maker, and I still want to run some tests on 2 things:

1. parallel vs. toe-in

2. simple Photoshop vs. StereoPhoto Maker

- Greg

 

Tobor

The toeing in is how the image may be adjusted for depth. This is something like what our eyes naturally do. You'd only have completely parallel vision when the eyes are focused to infinity.

What's (usually) a good rule of thumb is is to avoid parenting cameras in such a way that a lateral translation of the camera produces an absolute recentering of an object not at center. This causes the object to appear out of spatial context -- it has "zero depth" but objects around it may be confusingly forward or backward in z-space. This produces an "out of order" appearance, which is sort of what you have with your drawn example (based on other visual cues, other characters to your female look to be at the same plane, but in stereo view they are pushed forward or backward). You can set the toe-in stereo to be negative, positive, or zero parallax to your central character, so you need some method to set the depth.

That said, generally today toe-in stereo alignment is considered out-of-fashion (though, frankly, for a red/cyan anaglyph, who would really care much). The preferred approach is asymmetric (or non-symmetric) camera frustum, which at first glance looks like parallel cameras, but is not.

barbult

Tobor said:

The toeing in is how the image may be adjusted for depth. This is something like what our eyes naturally do. You'd only have completely parallel vision when the eyes are focused to infinity.

What's (usually) a good rule of thumb is is to avoid parenting cameras in such a way that a lateral translation of the camera produces an absolute recentering of an object not at center. This causes the object to appear out of spatial context -- it has "zero depth" but objects around it may be confusingly forward or backward in z-space. This produces an "out of order" appearance, which is sort of what you have with your drawn example (based on other visual cues, other characters to your female look to be at the same plane, but in stereo view they are pushed forward or backward). You can set the toe-in stereo to be negative, positive, or zero parallax to your central character, so you need some method to set the depth.

That said, generally today toe-in stereo alignment is considered out-of-fashion (though, frankly, for a red/cyan anaglyph, who would really care much). The preferred approach is asymmetric (or non-symmetric) camera frustum, which at first glance looks like parallel cameras, but is not.

How do you create "asymmetric (or non-symmetric) camera frustum" in Daz Studio?

barbult

Horo, I have created a new color anaglyph image where I have made sure the intersection with the border is at the screen depth. How does it look to you?

Tobor

barbult said:

How do you create "asymmetric (or non-symmetric) camera frustum" in Daz Studio?

Without a script or external software to adjust the images you probably can't. This topic has come up from time to time in the forums, and some folks have provided some ideas. We're not exactly in new territory here, fortunately. For example, found this thread in short order doing a G search:

http://www.daz3d.com/forums/discussion/46911/create-a-movie-in-3d-meaning-stereoscopic

Derek Carlin's comment seem pretty spot-on.

 

kaotkbliss

Would it not work to create a null object and place it where you'ld want the viewers eyes to be focused and have the cameras "point at" that null?

Take the above ballerina for example. If you place the null at her waist then everything behind it would be pushed back while everything in front of would be pulled forward.

Horo

barpuilt - fantastic how the dancing lady partly recedes and partly protruded. Very easy to see her in 3D. I'd say it's perfect.

 

barbult

Horo said:

barpuilt - fantastic how the dancing lady partly recedes and partly protruded. Very easy to see her in 3D. I'd say it's perfect.

 

Great! Thanks for checking it for me.

algovincian

barbult said:

Horo said:

barpuilt - fantastic how the dancing lady partly recedes and partly protruded. Very easy to see her in 3D. I'd say it's perfect.

 

Great! Thanks for checking it for me.

Pops nice, even though the parallax shift is minimal.

- Greg

barbult

algovincian said:

barbult said:

Horo said:

barpuilt - fantastic how the dancing lady partly recedes and partly protruded. Very easy to see her in 3D. I'd say it's perfect.

 

Great! Thanks for checking it for me.

Pops nice, even though the parallax shift is minimal.

- Greg

You guys are killing me with the technical terms . What does "the parallax shift is minimal" mean?

algovincian

I just meant the difference between the left and right image isn't that great, which, for me, makes it easy to focus and limits the ghosting. At the same time, your image still looks really 3D to me. It's an ideal situation - nice work!

- Greg

barbult

algovincian said:

I just meant the difference between the left and right image isn't that great, which, for me, makes it easy to focus and limits the ghosting. At the same time, your image still looks really 3D to me. It's an ideal situation - nice work!

- Greg

Thanks, Greg. I honestly don't know why the dancer seems to have minimal parallax shift and the moped rider had a lot. Both of them look like realistic 3D to me, when viewed though, not flattened or hyperstereo. I actually had the cameras slightly farther apart on the dancer image. Maybe it is because I used a gray background with the dancer so I didn't have to deal with the separation of both foreground and background. Maybe it is a difference in focal length or camera position. I have a lot to learn with rendering 3D in Daz Studio.

Edited to add: In both cases, the cameras were parallel with no toe in.

ramon73

barbult said:

Horo, I have created a new color anaglyph image where I have made sure the intersection with the border is at the screen depth. How does it look to you?

This is a very good anyglyph image! I am trying to create anaglyphs myself using this free software (https://anaglyph-maker.en.softonic.com/), but I am not getting the depth at all with my renders. When I move my head sideways then I do see movement in the image, but almost no depth at all. What does the camera offset need to be to get a good stereo perspective?.

andrushuk1

This was my first attempt :)