[DIY] Auto-Stereo with Parallax Barriers

Tutorials on how to create your own rigs, pics, movies, and everything that has to do with S-3D at home!
Post Reply
dukenukesam
One Eyed Hopeful
Posts: 23
Joined: Thu Jun 17, 2010 10:23 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by dukenukesam »

Thanks for the quick reply. What I meant was CAN I see the 3D effect on an interlaced image if I just put the Parallax barrier in front of it by doing it in Photoshop. I have done it and I don't get the 3D effect, it works in that it shows 2 positions angles for the image if I move the barrier one side or the other, but the Interleved image only looks 3d if I use shutters, I think the point is to do it without them.

I want to know if anyone has tried it since I would be able to come up with the barrier match beforehand in Photoshop and then just print the one I know will work.
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

If you are asking if this effect can be simulated in Photoshop, then no. You have to physically print out the barrier pattern and place it upon the screen for it to work.
dukenukesam
One Eyed Hopeful
Posts: 23
Joined: Thu Jun 17, 2010 10:23 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by dukenukesam »

No wonder it didnt wok lol. Thank you for your patience :D

Will give it a go with printing then.

Thanks!
User avatar
Likay
Petrif-Eyed
Posts: 2913
Joined: Sat Apr 07, 2007 4:34 pm
Location: Sweden

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Likay »

The print also needs to match the displays pixels perfectly to have stereoscopy through the whole screen. On onepixel basis i managed to have about 7 centimetres of good stereoscopy before the image doubled/switched. It was just luck that my printers native res was so close to the screens. Another way is of course increasing the pixelbase which of course will give more visible lines.
Mb: Asus P5W DH Deluxe
Cpu: C2D E6600
Gb: Nvidia 7900GT + 8800GTX
3D:100" passive projector polarized setup + 22" IZ3D
Image
dukenukesam
One Eyed Hopeful
Posts: 23
Joined: Thu Jun 17, 2010 10:23 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by dukenukesam »

Thanks guys. I actually have two 8" LCD's:

Pixels: 800(W) x 600(H) pixels
Active area: 170.4(W)x127.8(H)
Pixel Pitch: 0.213(W)x0.213(H)

I opened one up and put it in front of the other, can I make the second one show a barrier and have it work? I know printout is easier but was trying to have something like a 3d 2d monitor thing going on. :P

Can it work? Or maybe I just screwed up that second monitor lol. :lol:
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

dukenukesam wrote: Can it work? Or maybe I just screwed up that second monitor lol. :lol:
If the front monitor is translucent enough to show the back one, then I guess it could work. I know some prototype monitors have been shown using this method so it clearly works (at least using the right type of LCDs). This would also make alignment a little easier since its a 1:1 ratio. Although you will probably still experience the sub-pixel ghosting issues that I spoke about. The only way to fix that was the turn the monitor into portrait mode. Not sure if that would work using an LCD as the barrier.
SpAiK
One Eyed Hopeful
Posts: 5
Joined: Fri Jul 23, 2010 3:09 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by SpAiK »

Hello everybody!

I've been reading this full thread and I'm very interested in beeing able to convert my lcd into auto3d. My main interest is on lenticular sheets, but at first I would try with parallax barrier. So browsing the web I found someone that was succesfull on a 42" lcd with 9 views barrier, so how is this possible?
User avatar
android78
Certif-Eyable!
Posts: 990
Joined: Sat Dec 22, 2007 3:38 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by android78 »

SpAiK wrote:Hello everybody!

I've been reading this full thread and I'm very interested in beeing able to convert my lcd into auto3d. My main interest is on lenticular sheets, but at first I would try with parallax barrier. So browsing the web I found someone that was succesfull on a 42" lcd with 9 views barrier, so how is this possible?
I would say that he is using a bitmap mask that has one clear pixel column followed by 8 black:
BBBBBBBBWBBBBBBBBWBBBBBBBBWBBBBBBBBW
BBBBBBBBWBBBBBBBBWBBBBBBBBWBBBBBBBBW
BBBBBBBBWBBBBBBBBWBBBBBBBBWBBBBBBBBW
BBBBBBBBWBBBBBBBBWBBBBBBBBWBBBBBBBBW
BBBBBBBBWBBBBBBBBWBBBBBBBBWBBBBBBBBW
BBBBBBBBWBBBBBBBBWBBBBBBBBWBBBBBBBBW

The trouble is getting the 9 images required and generating the composite of these. I recently found the following which is a gif with 10 frames (well, it's more then that to get the complete cycle, but 10 unique images) which could be used:
http://www.gbeye.com/images/b2bproducts ... -480px.gif" onclick="window.open(this.href);return false;
If you get that image, increase the size by a factor of 9, overlay the above, and then splice the frames 1-9 with an offset of 1-9 pixels (depending on frame), it will create the picture to be displayed on the monitor.

This will, however, create a very dark picture and it may be better to use every 2nd frame and only a 5 view image (stereo viewable from 3 positions).
SpAiK
One Eyed Hopeful
Posts: 5
Joined: Fri Jul 23, 2010 3:09 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by SpAiK »

So this is limited to specially composed images, isn't it? Cause stereo rigs only record from two different views.

Anyway, he also mentions something about converting a lcd to 3d for use with polarized glasses, and I also read something about that in a study paper. Where could I find some more info to do that?
User avatar
Likay
Petrif-Eyed
Posts: 2913
Joined: Sat Apr 07, 2007 4:34 pm
Location: Sweden

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Likay »

The way to convert an existing lcd-monitor into 3d is using a quarterwave retarder (the result is circular polarization and the technology is called x-pol or u-pol). It's best done on industrial level though. :lol: Imagine yourself cutting all those stripes by hand and then manually attach them to the screen.....

As for android78 i'm also uncertain what software the author uses to make it work....

One way to make autostereo with parallaxbarriers would be just making one without matching the screen (which is very difficult if you don't have a very good laserprinter) and then make a calibrationsoftware on subpixellevel for the screen+barrier. Coloranomalies could also to some degree be adjusted away in the algorithm i think.
Mb: Asus P5W DH Deluxe
Cpu: C2D E6600
Gb: Nvidia 7900GT + 8800GTX
3D:100" passive projector polarized setup + 22" IZ3D
Image
SpAiK
One Eyed Hopeful
Posts: 5
Joined: Fri Jul 23, 2010 3:09 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by SpAiK »

What I've seen about polarization is as simple as putting a piece of cellophane covering one half of the screen. But the screen has to have polarization filter and also the glasses have to be adequate to the polarization angle that is output from the screen.

About parallax barriers, well, it's something I'll try if I find the exact values that have to be printed, then I'll go to a professional printing store to assure a precision printing.
User avatar
Likay
Petrif-Eyed
Posts: 2913
Joined: Sat Apr 07, 2007 4:34 pm
Location: Sweden

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Likay »

Lcd-monitors are already linear polarized. Applying a piece of cellophane (substitute for qw-retarder) doesn't make the screen 3d. If you cover one half of the screen with a retarder the light from the retarder will be circular polarized and the other linear. Linear polarized light easily goes through a circular polarizer and vice versa (means neither circular, linear or a combination of glasses will work). However if you apply two retarders on each half of the screen you can use circular glasses like real-d but in this matter you'd do just as well freeviewing crosseyed. I might misunderstood though.
Last edited by Likay on Thu Jul 29, 2010 6:14 am, edited 1 time in total.
Mb: Asus P5W DH Deluxe
Cpu: C2D E6600
Gb: Nvidia 7900GT + 8800GTX
3D:100" passive projector polarized setup + 22" IZ3D
Image
SpAiK
One Eyed Hopeful
Posts: 5
Joined: Fri Jul 23, 2010 3:09 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by SpAiK »

I may not have explained myself well, this is what I was talking about.
User avatar
Likay
Petrif-Eyed
Posts: 2913
Joined: Sat Apr 07, 2007 4:34 pm
Location: Sweden

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Likay »

Lol! and yes: This is simply a more complicated way of seeing crosseyed. :lol:
Many people would find it helpful though since each eye only sees the correct corresponding view. The polarizers and glasses simply helps blacking out the image that's not needed for each eye. Still you need to consider that the brain has to learn a new way of focusing, just like when trying to view crosseyed.
In my experience the cellophanes i've tried had more quarterwaveretarder-properties but practically worked out pretty bad. It's quite cool they found a fabric which have almost perfect halfwaveretarder characteristics.
I assumed that cellophane is closest to qw and that's why i promted two was needed (shifted 45° to the left resp 45° to the right). The final result using qw's is viewed with circular polarized glasses instead of linear.
Mb: Asus P5W DH Deluxe
Cpu: C2D E6600
Gb: Nvidia 7900GT + 8800GTX
3D:100" passive projector polarized setup + 22" IZ3D
Image
User avatar
android78
Certif-Eyable!
Posts: 990
Joined: Sat Dec 22, 2007 3:38 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by android78 »

SpAiK wrote:About parallax barriers, well, it's something I'll try if I find the exact values that have to be printed, then I'll go to a professional printing store to assure a precision printing.
I put the math for this up here:
http://www.mtbs3d.com/phpBB/viewtopic.p ... 7&start=42

Hope this works out for you. Let me know if you need help with the math.
fkiller
One Eyed Hopeful
Posts: 2
Joined: Thu Oct 14, 2010 12:20 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by fkiller »

I've just read you guys' great approaches, and wonder if any manufacturers develop parallax barriers film for specific device (e.g. iphone 4) and attach device that helps adjusting vertical pixels to match with film's LR position and an app that calibrates screen and demonstrate auto-stereos.
User avatar
Fredz
Petrif-Eyed
Posts: 2255
Joined: Sat Jan 09, 2010 2:06 pm
Location: Perpignan, France
Contact:

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Fredz »

I don't know about parallax barriers, but there is a lenticular sheet for the iPhone which is called 3DeeSlide which comes with a stereo image viewer available on the App Store (3DeeView).
See there : http://www.spatialview.com/product/3deeslide/" onclick="window.open(this.href);return false;
fkiller
One Eyed Hopeful
Posts: 2
Joined: Thu Oct 14, 2010 12:20 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by fkiller »

Fredz wrote:I don't know about parallax barriers, but there is a lenticular sheet for the iPhone which is called 3DeeSlide which comes with a stereo image viewer available on the App Store (3DeeView).
See there : http://www.spatialview.com/product/3deeslide/" onclick="window.open(this.href);return false;
This is great! Thanks Fredz! I need to buy it.
User avatar
Fredz
Petrif-Eyed
Posts: 2255
Joined: Sat Jan 09, 2010 2:06 pm
Location: Perpignan, France
Contact:

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Fredz »

You’re welcome. ;)
And don't hesitate to post a review if you happen to buy one...
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

Ok, I have revived this project yet again. Now I have finally gotten the full-screen implemented in standard landscape mode. Previously I was just using the monitor in portrait mode with windowed games. Now full-screen, full-resolution 3D gaming is a reality. I was able to play about 2-3 hours of Half-Life 2 with this setup, its is actually working. Yes, you heard me right. Glasses-free 3D gaming for under $1 dollar!!! I will be honest, I did get a little queasy by the end and I had to stop playing. That just shows how well it works. Now, it is not perfect. There is a significant, but acceptable, amount of ghosting going on. This is due to the way the LCD sub-pixels are arranged. If I use the monitor in portrait mode there is nearly no ghosting, but this is not really a viable solution for playing games. So ghosting is inevitable with this setup, but its probably not that much worse than, say, anaglyph. And no, it still doesn't match a real 3D monitor with 3D glasses, but it comes pretty darn close. Resolution-wise its about on par with the Zalman solution, although there is some graininess of the image you don't get on the Zalman. However in terms of brightness/color-reproduction it is actually better than Zalman (or maybe even shutter glasses). This is because you are not wearing any polarized glasses, which inherently tint the image (even the "transparent" side). So in terms of brightness/color quality, I found it to be brighter and have more natural color than even the Zalman 3D monitor. That is not to say the picture quality itself was better. With the parallax barrier it suffered from this graininess and increased ghosting, so it was still worse overall. Something like the Panasonic 3D plasma with 120Hz shutters will blow this away in terms of picture quality. However as a cheap DIY mod, this thing does wonders. The 3D effect itself looks pretty convincing and, if it weren't for the ghosting issues, grainy picture quality, and the pieces of tape in the middle of the screen, it could probably rival solutions costing $500 or more. I don't think I am going to get the quality much better than this. The problems I am facing are more of technical limitations rather than anything I am in control over. If I worked really hard I could probably get the ghosting down a little bit, but it is in a playable condition. The only thing I am trying to figure out at this point is some convenient way to attach and detach the barrier quickly (as of right now I have it taped at like 20 different points, it takes a while to put on or take off). If anyone has suggestions, please let me know. I am planning on putting together a full tutorial in the next couple of days, but let me just leave you with an image now so you guys can see the quality. Keep in mind that this image does not come close to the actual experience, but it is the best thing I could come up with. This is because the optimum viewing point is too close to the monitor to take a picture (although the video does a better job of representing the effect).

Cross-eye 3D image (can save as jps if you want):
ParallaxBarrier_cross_01.jpg
Here is a video of my testing image, just to illustrate the ghosting and viewing angles:
[youtube]http://www.youtube.com/watch?v=EstnDcN1Kwk[/youtube]

EDIT: Ok, so I tried to take some better stereo photos. Again, the ghosting is bad, but its not *that* bad. It looks way worse in the photos, not sure how I am going to document this. But these images should give you guys at least an idea of the quality.
ParallaxBarrier_cross_02.jpg
ParallaxBarrier_cross_03.jpg
ParallaxBarrier_cross_04.jpg
Also, the more I play with this setup, then more I am liking it. For some strange reason, the 3D effect almost seems more immersive than the Zalman 3D. I know that doesn't make sense, it should (at best) look the same. However something about it just looks really impressive. Maybe the dimming effect of the polarized glasses really makes that big a difference. I don't know. But its almost as if it looks more 3D. I'm not making this up.
You do not have the required permissions to view the files attached to this post.
User avatar
Okta
Golden Eyed Wiseman! (or woman!)
Posts: 1515
Joined: Tue Feb 12, 2008 5:22 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Okta »

Very cool. Maybe once you get the formula down you can a print shop to make a big one?
"I did not chip in ten grand to seed a first investment round to build value for a Facebook acquisition."
Notch on the FaceDisgrace buyout.
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

Okta wrote:Very cool. Maybe once you get the formula down you can a print shop to make a big one?
Yeah, I am looking into that now. Seems the biggest common size is 11"x17", which would almost fit my 22" monitor, but might be a little small for anything bigger. There are also large format transparency rolls, I just have to figure out if any local printers have them available. Even so, I am not entirely sure the pattern I have will work larger than 6" width. I seem to get ghosting along the edges, which is somewhat helped by having 3 sheets side-by-side (so I can adjust the phase manually). I have been trying to get a more accurate print, but I may have reached the limitation of 1200dpi, unfortunately. A print shop could have a higher-res printer but it will be difficult to go back and forth testing this to get the right pattern. Not sure if there is a better solution, though.
vpamir
One Eyed Hopeful
Posts: 7
Joined: Tue Dec 21, 2010 3:09 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by vpamir »

@cybereality

hi, i'm closely watching your work and i'm totally in love with it. i think the reason you've used 2 separate sheets at the left and right sides is to manually adjust them independent of the center piece to reduce ghosting, am i right? Well how about changing the pattern to reduce ghosting? the distance between the vertical lines of the barrier may be narrow in the middle and it may slightly grow when going to the edges. That may solve the ghosting problem but i think the driver must support this too. i don't think the vertical interlaced mode of iz3d presents any settings to compensate the new pattern layout but if it does somehow, maybe they add it as a feature in the future or even a parallax barrier mode which you can set the details, that maybe the way to go.

What do you think about my idea?

pattern example (think of it on a bigger scale with correct calculations):
pattern (reduced ghosting).jpg
You do not have the required permissions to view the files attached to this post.
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

@vpamir: Well, the reason I am using multiple sheets is because I do not own a large format printer. 8.5" x 11" is about the largest I can print, so multiple sheets are unavoidable. I initially thought the pattern had to change as it reached the edges (similar to your sample), but people here told me that is not the case. Supposedly the parallax barrier should have a fixed width for each line (ideally about the width of one physical pixel on the screen, or just slightly smaller). Anyway, I need this to be compatible with existing software, namely the iz3D driver but also the Stereoscopic Player, YouTube3D, etc. Its pretty useless if it only works with specially made content.

@Okta: I found that my local Kinkos will print on transparency, however they don't supply the sheets. I would need to buy them separately. The largest their printer supports is 11"x17", which could work for a 22" monitor. I can get those size sheets for about $1 a piece in a 10 pack (cheaper if I buy more). However the Kinkos printer is only 600dpi, which may not be detailed enough for this project. When I print at home at 600dpi, even with a working pattern (the same one used in the video/pics) it comes out weird and doesn't work. So I might have to find somewhere online that can do this at good quality (preferably better than what I have at home).
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

Ok, I have made a breakthrough. With this new method I have reduced ghosting to about 5% or less. After reading some whitepapers I realize the ghosting effect I was getting wasn't due to the pattern going out of phase, but rather sub-pixel bleeding at off-angles. If you are directly in front of the pixel everything is fine, but as you look toward the edge of the screen you start to see bleeding from adjacent sub-pixels. To fix this I am using 2 identical barriers layered one on top another. This has the effect of collimating the light coming off the screen and thus ensuring that light is emanating mostly perpendicular from the screen. This greatly reduces the sub-pixel bleeding I was seeing. Now it is not without some drawbacks. Because I am using 2 transparency sheets, the image is darkened somewhat (even just using 2 blank sheets does this, irrespective of the barrier printout). In addition, since light is being blocked at off-angles, this reduces the overall brightness of the image somewhat (although it was bad, ghosting brightness anyway). All in all, it is a marked improvement from the single sheet method. I still have some optimizations to do, and I think I can reduce the ghosting even further. However, where it is now in terms of ghosting is comparable to professional solutions.

You can see this illustrated in my latest video. The red/blue sample image I was using for testing was actually the worst possible thing I could have done, as it masks the sub-pixel bleeding issue (since red and blue are on the outside of the RGB pixel components). I still used it for the video, so you guys can compare to my earlier videos, but I am now using a simple black/white striped pattern which is way better for testing (you will see at the end of the video). Ideally the left image should be all black except the words "left" in white. The right image should be all white, except for "right" in black. You will see it comes pretty close, with barely any cross-talk on the text. It is very difficult to get pure white with this setup (its sort of tinted between yellow and cyan) but with a full-color image this is not so bad. Black is also not pure black, but fairly close. Hopefully I will be able to improve this somewhat.

[youtube]http://www.youtube.com/watch?v=IyrTVocRceo[/youtube]
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

So I have been doing some more experiments, but I don't think I am any closer to perfecting this. Basically I have left the period the same as with the last test video and adjusted the duty cycle. Basically this means changing the width of the opaque barrier lines in relation to the transparent spacing (without altering the period of the pattern). In the reference I have read, it was suggested a duty cycle of between 0.75 and 0.8 would yield good results. My previous attempts had been using equally sized opaque to transparent, or a duty cycle of 0.5. With a duty cycle of 0.75 you must increase the size of the barriers and you do indeed get reduced ghosting (in fact, ghosting is practically eliminated). The cost however is reduced brightness and some significant color inaccuracies. You see going with 0.75 that means each pixel is only being shown at half its width. When you take the sub-pixel RGB elements into account, this means only one sub-pixel component is being displayed at full brightness while the other two are being masked. Or, even worse, 2 sub-pixels are shown and one is completely occluded. This is disastrous. For example, if you want to display a medium gray color, it might be at RGB: 100, 100, 100. However if the red component is occluded you end up with a turquoise color, totally wrong. So this will not really work out. This is a direct result of having the RGB sub-pixels aligned horizontally. If I turn my monitor into portrait mode, then I do not get this problem. In that case a 0.75 duty cycle would work, however there is not much ghosting at all in portrait mode so it wouldn't even be necessary to do this. Then I tried reducing the duty cycle. I made a few quick attempts, and then went with a duty cycle of 0.346 for a more formal test. This has the advantage of increased light output (especially useful since using two overlapping parallax barriers causes a good deal of light to be lost). It also increases the accuracy of the color reproduction. Now colors are more vibrant and accurate than with a 0.5 or greater duty cycle (due to the sub-pixel masking explained earlier). However, this comes at the cost of increased ghosting. I was really excited to get near perfect color accuracy and about 30% more light, however the ghosting proved to be unacceptable. It also meant that in order for it to be playable, it was necessary to reduce the depth/separation so that ghosting was not as apparent. So while it was an interesting exercise, I might just leave the duty cycle at 0.5, as that seems to be giving the best overall results.

Here is a cross-eye image of HL2, using the duty cycle of 0.346:
ParallaxBarrier_cross_05.jpg


Notice the ghosting on the doorway on the back wall, on the CRT monitor, or the right edge of the desk. Especially on the edge of the desk notice how one eye is tinted red and the other blue. This is because the R and B components are on the edge of the pixel elements. This also gives the ghosting an almost anaglyph-like quality and greatly reduces the overall experience. For example, even the Zalman monitor will display some ghosting (under 10%) in certain scenes. However the colors do not shift at all, so this is easy to ignore. When the colors shift, you get into this retinal-rivalry situation, and it is not good. So these results are not even remotely acceptable. I would be much more inclined to sacrifice some light throughput or color accuracy in order to reduce (or eliminate) ghosting. So that is where I am at.

Also, maybe someone can help me with this:

The last major problem I am having with this project, is that I am getting strange banding around the edges of the parallax barrier printout. You can clearly see this on the right side of that HL2 image. The banding shows a rainbow effect of Red-Blue-Green, telling me that something is going wrong with the sub-pixels somewhere in the mix. At first I had assumed that this was because I had the period wrong. However I don't believe that to be the case. This is because if I adjust the barrier slightly to the left or right, the color banding continues to be on the edge of the paper. Also, if I move my head to the left or right, then the views themselves swap but the color banding remains on the edges. If it were a matter of the period being wrong and going out of phase, then the banding would follow me as I move my head. It doesn't. Even if I move the barrier to the edge of the monitor (instead of the middle) the banding is still on the edge of the paper (not in the same place it was physically on the monitor). So this tells me that it is not related to anything physical on the monitor. Strangely enough, if I print a small barrier (like 4" by 4") then the banding is on the edges still. With the same pattern printed out as 8" x 8", there is a good 6" x "8 area in the middle that is perfect, with the last 1" on each side having the banding. My only guess is that my printer somehow is not as accurate along the edges or that the laser possibly has some ramp-up/cool-down phase which makes only the middle of the paper pixel perfect. I could, I guess, print a full page barrier and then crop 25% off of each side. This might work, but then I am wasting resources (ink and transparency film) not to mention the increased tape needed to mount it on the monitor, it begins to get extremely ghetto. I will have to find a solution to this, preferable something more technical than just physically cropping the sides. This is the last real unknown in the equation, most of the rest of it I have down to a science. Its not going to be much longer before I am able to get this fully functional.
You do not have the required permissions to view the files attached to this post.
User avatar
Okta
Golden Eyed Wiseman! (or woman!)
Posts: 1515
Joined: Tue Feb 12, 2008 5:22 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Okta »

Could the banding be the spreading of the light spectrum because of the angel you are looking at it? Does it do this if you move your head to look straight at the area?
"I did not chip in ten grand to seed a first investment round to build value for a Facebook acquisition."
Notch on the FaceDisgrace buyout.
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

Okta wrote:Could the banding be the spreading of the light spectrum because of the angel you are looking at it? Does it do this if you move your head to look straight at the area?
That is what I initially thought, and I did have that effect happening over the whole sheet. However, after I added the second parallax barrier layer this should not be an issue anymore. This is because the light is being collimated, so very little light is escaping from undesirable angles. In addition, the edge of the barrier is always wrong. No matter where it is positioned on the screen, or where I position my head. If it were simply a matter of the angle of incidence, then if I moved my eye directly in front of the errant area, it should then become clear. It doesn't. It continues to have color banding now matter what angle I look at it from (the colors do shift somewhat, and ghosting can get better or worse, but it never has a clear view of any single frame). At this point I am almost sure it is something physical happening with my printer. I am going to do some tests today to see if I can verify this. If need be, I will buy a new printer. I really want to finish this project proper.
User avatar
Fredz
Petrif-Eyed
Posts: 2255
Joined: Sat Jan 09, 2010 2:06 pm
Location: Perpignan, France
Contact:

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Fredz »

I'm not sure changing printer will give better results. Without knowing what's the real cause of this problem, at first sight it sounds logical that the edge of a transparent sheet could be more deformed than the center (because of heat probably) because there is less "tension" at the edges. In the center the "tension" should be quite equilibrated since there is the same amount of the same matter (plastic) on both side, but that's not the case on the edges. Just speculation on my part after some (lot of ?) fine bottles of champagne though, so take this with a grain of salt... ;)
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

@Fredz: Yes, you are right. It could very easily be something with the paper itself, and this almost seems like the obvious answer. Either the way they are manufactured or some interaction with my printer (ie unevenness of the heating element of the laser printer). I don't really know, but it is something to explore. I guess I could buy some different brand transparency film (maybe some that is slightly thicker) and see if I get better results.

I also did a few tests today and I have verified that it is not an issue of the actual pattern design. I think I have the correct pattern (not 100% perfect, but good enough). What I did was print a small barrier square at 4" x 4" in the center on the paper. When testing it, it looks perfectly fine. Then I printed the exact same 4" x 4" file, but in the top-right corner on the paper. When I test that it shows the strange color banding / ghosting on the side close to the edge of the paper. So it definitely has something to do with the edge. However is it a problem of my printer not being accurate on the edge, or is it a problem of the paper. After reading up on how laser printers work, it does not seem likely it would have trouble on the edges (as I was previously thinking it worked more like inkjet printers do). It seems the mirror wheel (which directs the laser light) spins at a constant rate, there is not any stopping and starting on each raster line like with inkjet. So my ramp-up/cool-down theory seems to be bunk. However I am seeing something weird that may or may not be related. Basically, every printout I make of the pattern is not the same. I have maybe 10 prints (at different sizes) of the same exact pattern, but they all look slightly different. Some work really well, some not so much. They are mostly in the same ballpark, but it makes me uneasy. I did a lot of tests with small differences in the variables, but now I am not sure if any of that data is valid, if the printer cannot print the same file and have it come out accurate two times in a row.

Also, a new printer could help me out a lot. First off, I was looking at getting a semi large-format inkjet printer w/ 1440dpi. So right off the bat its all new technology, so I could try with inkjet transparency film and a slightly higher dpi. Not sure if that would be better or worse (I imagine better), but I would certainly have more to work with. The printer I am looking at can support a max 13" x 44" so even if I had to crop a foot off each side, I could still cover my full 22" monitor with one sheet. Of course, this would not be cheap, for the printer or for getting transparencies in that size. However it could be done.
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

Update:
So I went ahead and got that printer I was talking about. I've got it right now, but its still in the box since the transparency film didn't come yet. This printer can do 5760 x 1440 dpi with a minimum ink droplet size of 3 picoliters (diameter of a human hair would be equivalent to 12 picoliters). So I think I am good on the resolution. My current printer is only 1200 x 600 dpi, which is limiting. Beyond the "low" resolution, my printer is also giving inconsistent results. If I print the same pattern but alter the left margin slightly (say from 0.2" to 0.3") then the results vary significantly. So my current printer is just not up to the task. I could probably find some magic set of numbers that will work, but I am hoping that this new printer will just solve all these problems. I really hope it works out, not sure if I can afford any better printer than this one (they start to get expensive when you need wide-format and such high resolution).

Also, I did some experimentation with using a checkerboard pattern instead of the vertical interleave. So far it didn't work that well (because the 600dpi limit) but it did seem to have some effect. With this new printer I am hoping this will be possible. I don't think the checkerboard will give a better 3D effect, or solve the ghosting issues, but it does seem to give the appearance of a higher resolution. I will have to do some more tests to verify this, but it looks promising.
Last edited by cybereality on Sat Jan 08, 2011 9:39 am, edited 1 time in total.
vpamir
One Eyed Hopeful
Posts: 7
Joined: Tue Dec 21, 2010 3:09 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by vpamir »

I'm waiting impatiently :woot
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

So I started working with the new printer I got. I think its pretty much just what I needed, its leaps and bounds better than what I was working with before. I am really within spitting distance of having this 100% functional. It makes everything I have done up until this point look like a proof of concept. Now I have something that is actually playable and quickly approaching commercial quality. Though due to limitations of consumer printers, it will never match a real commercial made parallax barrier system like the Sharp display in the Nintendo 3DS. However what I have now is unquestionably better than anaglyph, and could even compete with interlaced solutions like the Zalman. In some ways it looks even better than glasses-based solutions because you are not losing any color due to polarization. You are still halving the resolution, but the color looks just like normal (aside from the sub-pixel ghosting, which I am working on). With polarized glasses like the Zalman, or even shutter glasses, you are always darkening the image and muting the colors. With the parallax barrier you don't have this problem, so in some aspects it can even look better (ie more vibrant).

Its not perfect, though. You still get this grainy image quality, and banding due to the interpolation used to fake a higher dpi on the print. There is not too much I can do about this unless I had a printer with like 30,000dpi. And the viewing angles are not so hot. There is only a specific spot that is perfect, otherwise you will get ghosting. However in that specific spot the ghosting is very low, maybe only 5%. I still need to optimize the pattern a bit more, but I am extremely close at this point. Its at the point where I am dealing with micrometers of inaccuracy. Should have the pattern finalized by this weekend. Then I just have to figure out how I am to document all this. If I had like the Fujifilm W3, it might be a little easier to take 3D photos. Right now I am just using a regular camera and moving it to the side, which isn't particularly accurate. Plus the camera does not capture what I am seeing in real-life at all. I have posted two cross-eye shots below, but I assure you the experience looked way better. Keep in mind that this is on a small 640x480 portion on the display (not full-screen) so that is why it looks so low-res. I should have better luck taking a 3D video, maybe I will try that tomorrow. For now you guys can at least see that the quality is much improved from my old printer, and that there is none of the color banding I was getting before. Please, let me know what you guys think.
DIY_Parallax_Barrier_2011_01.jpg
You do not have the required permissions to view the files attached to this post.
Last edited by cybereality on Wed Jan 12, 2011 2:05 am, edited 1 time in total.
User avatar
Fredz
Petrif-Eyed
Posts: 2255
Joined: Sat Jan 09, 2010 2:06 pm
Location: Perpignan, France
Contact:

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Fredz »

Doh!, that's pretty good considering you are using a consumer printer. It would be nice if you could post a shot of the DDD ghosting test and your blue/red test picture so we can better appreciate the result and compare it with other stereo solutions.

I guess your solution could also be used on mobile phones if their DPI is not too high, did you try to make some tests with this kind of device if you own one ?
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

Fredz wrote:It would be nice if you could post a shot of the DDD ghosting test and your blue/red test picture so we can better appreciate the result and compare it with other stereo solutions.
Yeah, I can totally do that. Still have to refine the pattern a bit, but I can do some preliminary ghosting tests tomorrow.
Fredz wrote: I guess your solution could also be used on mobile phones if their DPI is not too high, did you try to make some tests with this kind of device if you own one ?
Yes, that is in the roadmap. I want to finish the monitor mod first, as I am so close to having it complete.
User avatar
Fredz
Petrif-Eyed
Posts: 2255
Joined: Sat Jan 09, 2010 2:06 pm
Location: Perpignan, France
Contact:

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Fredz »

Nice, keep up the good work... :)

I wish I'd been equally dedicated to my projects as you are, some nice things you're coming with lately.
User avatar
Likay
Petrif-Eyed
Posts: 2913
Joined: Sat Apr 07, 2007 4:34 pm
Location: Sweden

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by Likay »

Now you're really making progress!
I lately just got my w3 camera and just started to play with it. One of my fave things is actually the display. Even if it flickers a bit i really appreciate the image from it from coming from such small device. I actually have second thoughts about autostereoscopic tv's and their hotspots being acceptable if several viewers can sit in each dedicated hotspot. The w3 looks best viewed straight front but the image is acceptable at two sideviews as well. So practically it's 3 hotspots with proper stereo with this camera.
Implemented on a tv-set with a good quality with oppurtunity to grant each viewer his own hotspot i can't figure it's really a bad idea. And that's coming from me....
(note though: The tech on the w3 isn't parallax barriers but is practically experienced in a similar way)
Sorry for the offtopic rant but i'm actually developing a bigger acceptance for these kind of techniques and you're doing quite some succes!
Mb: Asus P5W DH Deluxe
Cpu: C2D E6600
Gb: Nvidia 7900GT + 8800GTX
3D:100" passive projector polarized setup + 22" IZ3D
Image
User avatar
android78
Certif-Eyable!
Posts: 990
Joined: Sat Dec 22, 2007 3:38 am

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by android78 »

Great progress there!
I was curious if you are still using the monitor rotated to 90 deg. with the test pictures you uploaded? I ask because I see right eye has red ghosting and left eye has blue ghosting. If so, what is it like when you do have the monitor rotated to 90 deg?
Does anyone know if there are any LCD monitors that have the sub-pixels stacked vertically as opposed to the usual horizontal?
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

Thanks for the support, guys.

@Likay: Yes, there can be multiple zones with this DIY solution too. If you move to the side it gets blurry and pseudo-stereo, but if you move a little further it will become in 3D again (but of lesser quality than the center zone). So it could technically be possible for 3 people to view it, however the way I am doing it you have to be really close to the screen so it would be hard for people to be that close together. I mean, if they were all sitting on a bench directly next to each other it could work, but there is not as much room for 3 full-size computer chairs. For my purposes, single-user is fine because this is mainly for playing PC games which are a solo affair.

@android78: No, I am using the monitor in standard landscape mode. Although the portrait mode gives significantly better results, it is not really an option since most games do not support it. In addition you waste about 60-70% of the screen real-estate, so it is not really an immersive experience (which is the entire point). I did look for monitors with a vertical RGB arrangement, but I do not think they make them (at least not in consumer displays, maybe for professional auto-stereo panels). Anyway, I think the quality is acceptable, even with some minor sub-pixel bleeding. I don't want to call it ghosting necessarily, because you don't really see the full image from the other eye. Meaning you do not lose the stereo effect, its just that the colors are not accurate. I can probably reduce this further, but its an inherent limitation with having the RGB elements arranged horizontally. There is only so much I can do.
User avatar
cybereality
3D Angel Eyes (Moderator)
Posts: 11406
Joined: Sat Apr 12, 2008 8:18 pm

Re: [DIY] Auto-Stereo with Parallax Barriers

Post by cybereality »

So here are some more images. Not 3d, just ghosting tests. After testing tonight and playing some more of UT2004 and HL2, I realize the ghosting is worse than I thought. Its probably closer to 15-20% actually. Light on dark is the worst, so this is a lot more apparent on HL2 than it was on UT2004. I mean, its still in a playable condition and the 3d effect is nice. But the ghosting can be a bit distracting. I can probably reduce the ghosting by a good amount by using two barriers, but this also has the effect of reducing light and altering the colors. I might try that next. Anyway, here is where I'm at:
DIY_Parallax_Barrier_2011_03.jpg
DIY_Parallax_Barrier_2011_04.jpg
DIY_Parallax_Barrier_2011_05.jpg
You do not have the required permissions to view the files attached to this post.
Post Reply

Return to “Do it Yourself!”