[DIY] Auto-Stereo with Parallax Barriers

Post by **angelsinraver** » Sun Dec 21, 2008 1:38 am

hey is anyone still working on this? it seems really interesting. think ill try a test.

Post by **Likay** » Sun Dec 21, 2008 2:54 am

I tried. It seems like my monitor dpi size as well as the laserprinter dpi size almost matches which gives me a "proper" effect. It only works for a few cm's before the pixels don't match any longer though. Not keen on getting another printer or increase the pixelsize so i gave up on it. It's absolutely possible but if you want autostereoscopy on pixelbase you're going to need a very good laserprinter. The foil i was using is foil for laserprinters and seems to keep the shape without problem (during printout the sheet passes through a heating drum in the laserprinter for hardening the toner purpose).
The sheets i used was "Corporate express: Overhead transparencies: A4: 608 83 29". This foil seemed to keep the shape through the laserprinter and gives no problems.

cheers

Post by **Okta** » Wed Jan 07, 2009 7:09 am

Do a google for lenticular screens. It seems there are many companies out there who sell lenticualr sheets of all shapes and sizes and custom jobs. Would be cool if you can find a matching one and just stick it over your lcd. Instant autostereo using interlaced driver.

Post by **LukePC1** » Wed Jan 07, 2009 1:51 pm

problems:
- highly acurate barrier (plastic might stretch when heated up)
- acurate positioning
- color and positioning of subpixels (for LCD)
- blurryness of CRT (but they can be rezised well

Unfortunatly they have a curved scren

- maybe a barrier for a very low resolution works. A 1280 LCD could be driven at 640 e.g.

I wish best luck

Post by **WebDext** » Wed Jan 07, 2009 7:26 pm

What's interesting about the RGBRGBRGB aspect of LCD monitors could actually be used to it's advantage I think. If using head tracking you can determine where the ray locations from each eye hit the screen, as well as the projected width of the vertical strip on the monitor, you could calculate all of the Red Green and Blue sub pixels that fall within that projection.

Then all you would have to do is create an array of groups of sub pixels for each eye, and map the color values to those sub pixels using linear interpolation.

So say for example you know the first pixel of the right eye is going to see a group like this:

|RGBRG|

You now have a rough equation for the color at that location: 2*R+B+2*G
Break that up into components and set it equal to the determined visible color at that location (probably the first color for the right eye)

2*Rscreen = Rimage or Rimage/2
etc..

Additional weighting of brightness values could be calculated as well, and the algorithm could be fairly well optimized because you only have to solve the weighting once for the columns then you can apply it to all the pixels in that column.

The color distorted output image should then look correct for each eye, and you've taken advantage of the higher horizontal resolution.

Any thoughts?

Post by **jem454** » Tue Feb 17, 2009 4:37 pm

Very helpful lenticular info.

http://www-isl.kaist.ac.kr/Papers/IC/ic117.pdf

Post by **3dfreak** » Sun Apr 19, 2009 1:44 pm

i realy enjoyed reading this thread:

And i might have an idea that I want to try:

I want to use this 7 Inch screen with a resolution of 800x480 pixel.
http://www.geekstuff4u.com/accessories/ ... splay.html" onclick="window.open(this.href);return false;

Here are some specs: Display size: 6" × 3.6" (15.25cm × 9.15cm) = 133.28 PPI with a side ratio of 4 : 2.4

I calculated that it has approx. 30 lpi.

Now I thougt about buying a lense with 30 lpi and putting it on the screen.
Could i use the iZ3D driver on Interlaced mode or the stereo photo maker on interlaced output to see some pictures?

Could it work?

Post by **Likay** » Sun Apr 19, 2009 2:54 pm

Theoretically yes. Practically... let's see.
It's all about how well the lenticular lense matches the display. Please come back with results if you go for it. I'm curious.

Post by **3dfreak** » Mon Apr 20, 2009 12:29 am

does anybody know a small display with a small resolution of 320x240 with a sice of 7" or simillar? It might be a lot easier with a smaller resolution but same sice.

Post by **3dfreak** » Thu May 07, 2009 4:40 pm

well I tried the 3Dee Shell from Spatial View:
http://www.spatialview.com/en/node/489" onclick="window.open(this.href);return false;

It turns my iphone into an autostereoscopic screen using a lenticular lens. It is pretty amazing actually. The 3d effect is impressive for this rather small device. I will post some pictures later and maybe a review of some kind.

greetings and keep working on the autostereoscopic screen, i dont see any movement here, keep it up guys!!!

Post by **cybereality** » Tue Sep 29, 2009 10:14 pm

So I have resurrected the parallax barrier project. I got a 1200dpi mono laser printer, the Samsung ML-1630 ( http://www.overclockersonline.net/?page ... s&num=1742" onclick="window.open(this.href);return false; ) and a 50 pack of laser transparency film from Sparco. So far I have been getting good prints with the ML-1630, it is definitely more precise than that old 300dpi inkjet I was using last year when I started this. At 600dpi I was able to print a 5 pixel interlace pattern (5px black, 5 px blank, etc.) and it was nice and sharp with no bleeding or anything.

I did some math and I figured for a 22" 1680 x 1050 display the individual barrier lines should be around 6.7 pixels @ 600dpi. So I tried with a 6px line and also a 7px line @ 600dpi and the results were close but not there yet. It seems 600dpi wont cut it. Then I tried with a 13px / line @ 1200dpi and I am getting somewhere. I still have a lot to figure out, but I am at least in the ballpark here.

The image below is with the 13px barrier and on the screen is a 1px vertical interlace pattern of blue and red. If the parallax barrier worked perfectly then you would be seeing either total blue or total red on the screen. As you can see there are some strips where it does block out the intended pixels and you do see either red or blue. However it goes out of phase and the effect is lost. But I am getting somewhere, much closer than I got with that old inkjet.

ParallaxBarrier_01.jpg

Since I am at the maximum dpi my printer can handle I will need to figure out some techniques for increasing the precision. One method would be to keep the lines at one sizing (say 13px) but alter the spacing (to say 11px) to offset the phase. If that is not enough I may have to create a more complex pattern, where say every 10 lines it adds or removes some pixels to offset the pattern. Using a combination of these methods I think I can attain the proper 1:1 pixel mapping necessary. So far this has only been with 2 days of experimenting so I imagine I could have something working in a couple weeks maybe. Stay tuned.

Post by **Okta** » Wed Sep 30, 2009 1:27 am

Wouldnt it be easier and cheaper to take your image on disk to a print shop and get them to print onto your film with their higher quality printers? Should only cost a few bucks max.

Post by **Likay** » Wed Sep 30, 2009 11:14 am

Okta wrote:Wouldnt it be easier and cheaper to take your image on disk to a print shop and get them to print onto your film with their higher quality printers? Should only cost a few bucks max.

Sure, but you might have to make 5 prints to find out about which resolution that suits best and then maybe 5 more to finetrim it. I don't think he bought the printer just to catch up on parallax barriers though.

Nice test cyber. What happens if you try to make a parallax barrier invertical direction instead? (you do have to tilt the monitor too. I had best "success" doing this).

Post by **cybereality** » Wed Sep 30, 2009 6:24 pm

Okta wrote:Wouldnt it be easier and cheaper to take your image on disk to a print shop and get them to print onto your film with their higher quality printers? Should only cost a few bucks max.

Not really. I needed a new printer anyway and it was on sale for $90. Plus the transparency film is only like $10 for a 50 pack. And it is much easier for me to push a button and then 3 seconds later have something in my hands rather than burning a CD (15¢), taking a train to a copy shop ($5), waiting around for them to print it (time=$), paying their rip-off prices for each print ($$$), etc. Once I get the template finalized I will be sending it off to a professional printer so I can get a full-sized print that will fill the whole monitor. But I can't be going to the copy shop every time I need to make a sub-pixel adjustment.

Post by **cybereality** » Wed Sep 30, 2009 9:43 pm

Ok, so I have made some more progress. I am much closer than I even thought. You were right, Likay. I needed to turn the monitor into portrait mode. I already knew I was going to have to do this eventually, but I thought since I was only testing with primary colors (red/blue) that I could avoid the sub-pixel striping issues. Apparently not. So I am getting much better results in portrait mode. I also altered the pattern so it alternates between 13 px lines but every 6 lines it inserts 1 blank pixel. This is a lot closer to what I need, I am almost there:

ParallaxBarrier_02.jpg

ParallaxBarrier_03.jpg

Post by **cybereality** » Sat Oct 03, 2009 6:57 pm

Ok, so I have made some more progress. Right now I am using a base of 13 pixel wide barrier lines @ 1200 dpi but every 6th line is 14 px and then every 24th line is 15 px. This is extremely close to what I need but I am still off by some small fraction of a millimeter. This is giving me some ghosting and its losing the phase at the edge of the sheet. However it is like 80% there and I don't think it will be much longer until I have it working 100%. It is certainly working good enough as a proof of concept and I am sure at this point that the method is sound. I have recorded a short test video just to illustrate the effect. Its not perfect yet, but you should see it is very close. The video is uploading right now, will post in a minute.

Anyway, I was able to play about a half hour of Mirror's Edge using the parallax barrier and it does work, I was getting a nice 3D effect. That said, there was pretty significant ghosting and the viewing angle was painfully narrow. In fact, I wouldn't even call it a "viewing angle" it was more like a single point in space. Move outside the point and there was either a lot of ghosting or the eyes would swap. However within the "sweet spot" the stereo effect was legitimate and looked decent. There was also a graininess that the barrier added making the image quality somewhat muddy, but otherwise it was full color 3D. I won't be selling my Zalman any time soon, but all things considered this DIY mod came pretty close for what amounts to 25 cents worth of a transparency sheet. Stay tuned.

Post by **cybereality** » Sat Oct 03, 2009 7:37 pm

[youtube]http://www.youtube.com/watch?v=g5V_CKSlpXc[/youtube]

Post by **Okta** » Sat Oct 03, 2009 7:46 pm

Lol that is awesome

Post by **cybereality** » Sun Oct 04, 2009 11:57 pm

Right now I am in spitting distance of having this thing working proper. It is already somewhat playable. I have fine-tuned the pattern a bit and I am now adjusting within the hundredth of a pixel. It won't be long before I find the right number. Tonight I was able to get rid of some of the ghosting from the last test. It may look the same, but if you watch the video closely you will see there is slightly less cross-talk.

[youtube]http://www.youtube.com/watch?v=gRZ59HfV2QY[/youtube]

I also attempted to take some stereo photos of the setup, but they didn't come out too good. I was just using my standard Canon PowerShot and offsetting the shots a bit. You can kind of get an idea, but it really doesn't capture how it looks in real life. For me to do this proper I will need that Fujifilm W1 and a tripod (neither of which I have at the moment). But this should be enough to at least prove I am not full of it.

DIY_AutoStereo_01.jpg

DIY_AutoStereo_02.jpg

Post by **Likay** » Mon Oct 05, 2009 10:18 am

Awesome work!

Post by **cybereality** » Mon Oct 05, 2009 10:29 pm

Using the same pattern from last time I have printed it out a little bigger and took some more photos. I also attempted to shoot some video with my dual-Vados but I don't have the time tonight to process any of it. It was very difficult to do this hand-held but hopefully I captured something. Anyway, check out the shots:

DIY_AutoStereo_03.jpg

DIY_AutoStereo_04.jpg

DIY_AutoStereo_05.jpg

Post by **Okta** » Tue Oct 06, 2009 3:44 am

Do you see the same vertical blind effect with your eyes or do the stereo images smooth it out some?

Post by **cybereality** » Tue Oct 06, 2009 8:37 pm

Okta wrote:Do you see the same vertical blind effect with your eyes or do the stereo images smooth it out some?

Well those are cross-eye images so try them out for yourself. You can also download them and rename to "JPS" and open with a modern solution.

Post by **Okta** » Wed Oct 07, 2009 5:08 am

Yeah its still there, but that could be classed as an acceptable gotcha by many for such a solution. So next step is the driver, were those shots taken using an existing interlaced driver or did you custom create the stereo pairs as still shots?

Post by **mickeyjaw** » Wed Oct 07, 2009 2:00 pm

Okta wrote:So next step is the driver, were those shots taken using an existing interlaced driver or did you custom create the stereo pairs as still shots?

Dunno, but i've just figured out how to use mplayer (windows or linux, possibly OSX too) to convert to and display vertical interleave on the fly from any type of 3D input format i.e over/under, left/right, field sequential etc. If anyone is interested, the thread is here:

http://www.mtbs3d.com/phpBB/viewtopic.p ... 21&start=0

Also, it should be possible to output 1px, 2px, 3px, 4px etc vertical interleaves without compromising the horizontal resolution using this method, give me a shout if anyone needs this and I will post further instructions

Post by **cybereality** » Thu Oct 08, 2009 7:09 pm

Okta wrote:So next step is the driver, were those shots taken using an existing interlaced driver or did you custom create the stereo pairs as still shots?

Thats already taken care of. I'm using the IZ3D driver in vertical interlace mode. So I was actually playing Mirror's Edge for like a half hour with the parallax barrier and when I found a good part I would just stop and take a snapshot.

Post by **Okta** » Thu Oct 08, 2009 7:24 pm

Sweet. How did you find the experience? Would you recommend this over say anaglyph? Im getting keen to print up a big sheet for my 28inch

Post by **cybereality** » Thu Oct 08, 2009 11:32 pm

Okta wrote:Sweet. How did you find the experience? Would you recommend this over say anaglyph? Im getting keen to print up a big sheet for my 28inch

I thought it looked decent for what was basically a 25 cent piece of transparency paper. Still a lot of issues I need to work out, so don't think I am finished just because I have something playable. It certainly looked better than anaglyph when it was working, but due to a variety of issues its not yet at the point where I could recommend it to anyone. Its still a work in progress.

Post by **Freke1** » Fri Oct 30, 2009 1:00 pm

Found this:
[youtube]http://www.youtube.com/watch?v=A11_M4LqL_8&feature=sub[/youtube]
looks good when he moves the film with his fingers.
How nice it would be if You could just buy some film to stick on Your monitor...

Post by **cybereality** » Fri Oct 30, 2009 5:56 pm

Freke1 wrote:Found this:
...
looks good when he moves the film with his fingers.
How nice it would be if You could just buy some film to stick on Your monitor...

Wow! Nice find. Would be hard to sell lenticular kits since each panel has a different dot pitch it would be insane to support all the different displays. But if people could find a cheap monitor that works well with existing lenticular sheets (like that guy apparently did) then it wouldn't be too crazy to simply buy the monitor that worked with the lens.

Post by **cybereality** » Sun Nov 01, 2009 11:44 pm

I've posted a new video. Finally got that tripod so now I can record and play at the same time.

Still not complete yet but playable:

http://www.youtube.com/watch?v=goPFfIBb9aI" onclick="window.open(this.href);return false;

Post by **android78** » Thu Dec 03, 2009 6:43 pm

Well done Cyber!!!
Just a couple of questions (interested in a very big project for 52inch TV):
1. Are you still using the monitor in portrait mode, or did you get it working in landscape?
2. is the interlacing on the screen just every second column of pixels, or is it using multiple columns of pixels?
3. Is the barrier layer sitting flat against the screen glass, or is there another layer to make it further from the screen?
4. Have you considered using wider lines for the black to get better viewing angle? I'm aware this will make the image darker, but it might make it better to use.
so have the columns you are printing:
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
5. I'm considering doing a checkerboard so that I can play Avatar on xbox360 as:
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
BWWBBBBBWWBBBBBWWBBBB
BWWBBBBBWWBBBBBWWBBBB
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
BWWBBBBBWWBBBBBWWBBBB
BWWBBBBBWWBBBBBWWBBBB
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
BBBBBWWBBBBBWWBBBBBWW
What are your thoughts?

Post by **cybereality** » Thu Dec 03, 2009 11:22 pm

Yeah, I am using the monitor in portrait mode. Landscape cause some pretty bad sub-pixel ghosting. There may be a way to fix this but it would limit the type of content that could be viewed. The interlace pattern I am using blocks every other pixel, so it is a 1 pixel barrier. I have it taped directly to the monitor. It seems the glass itself generates enough parallax for it to work. I have considered making the lines thicker to increase viewing angles. I haven't explored this idea much, but it should work in theory. I still want to work on this project more but I keep getting distracted. Doing a checkerboard pattern would be a lot more difficult but I guess it could be possible.

Post by **android78** » Mon Dec 14, 2009 3:45 pm

I have managed to create a parallax barrier for the sony PSP (3000, not Go).
I'm not sure if the scaling of the printer I've been using is out, or if there is some other issue, but I couldn't just set the resolution as I expected. What I found worked is using GIMP and set the resolution to 128.8 pixels per inch (for anyone not using photoshop, I recommend this application as it's one of the best open source applications out there). This creates a barrier that is almost perfect for the psp when held at arms length.
I printed on cheap transparency paper on a laser printer at work, and there was a little smudging of the toner, but good blockout from the printed lines.
There are some issues though:
1. There is about 2mm between the LCD and the plastic in front of the screen on the psp. This means that for optimal viewing, you need to hold it at about 2 feet from your face.
2. The psp screen is a bit too dark for this to work effectively. I converted one of the screen shots from avatar that Neil had uploaded, and had to make the brightness 150% before it was really viewable.
3. Obviously of no fault of the psp or this solution, but a general fault with parallax barrier solution with only two images, but you must hold your head very steady as shifting more then a couple of cm either way will cause ghosting. More then about 3 cm (half way between your eyes) will reverse the image causing headaches. One thing that I will try is with a 2/3rds barrier and 3 images when I get a chance. my calculation indicates that this should work well on the psp at about 12 inches from the screen, so long as it isn't just too dark to see anything.

All in all though, I'm pretty pleased. I would love to see a game to implemented vertical striped L/R stereoscopic views on the PSP.

Where to from here:
1. See if I can implement 2/3 barrier with 3 images.
2. See if I can create barrier for something bigger... checkerboard for 52" screen

Post by **Erant** » Mon May 03, 2010 2:09 pm

I've recently become interested in the parallax barrier way of creating a cheap, simple (ish) 3D display. I stumbled across this forum in my research and figured I'd share with you what I have so far. I hope you'll forgive me for resurrecting such an old topic, but I thought this post belonged here. I've created a small C program that generates a parallax barrier for you, based on the equipment you have at hand. This includes the DPI of the printer you're using, the pixel width of your screen, the width in pixels of your display, and the height of your display. My program will then generate the optimal barrier for your display.

This method isn't perfect, it doesn't incorporate the possible stretch of your transparency sheet, nor the minor imperfections in your printer. After toying around with it for a while, I found a 0.2295mm dot pitch to be the closest to my MacBook Pro 15" screen, which at 1440x900 has a .231mm dot pitch. It still isn't perfect (I can make an image of about 12cm wide appear in 3D), which I'm attributing to the fact that this dot pitch just happens to be very incompatible with 300, 600 and 1200DPI printers. For this reason, my program also outputs some statistical information, to help you assess the quality of the produced image.

I'm still working out a good way to determine what you should do to modify your barrier to be pixel perfect. I'm currently working on the math of alternating-pixel-images, which show more or less banding when viewed with one eye depending on the quality of your barrier (more banding means a more incompatible barrier).

The code uses libpng to create the output image (which can become quite large, in the range of 10k pixels x 10k pixels). You must print this at the DPI you specified when generating, otherwise it won't work. Because it uses libpng, you must also link to it when compiling. On my Mac, this boils down to something like:

Code: Select all

gcc -I/usr/X11/include -L/usr/X11/lib -lpng -O2 -std=c99 main.c -o parallax

Then run ./parallax and read the instructions. (Example commandline, for my Mac: ./parallax 1200 0.2295 180 1000 macbook-example) This should work fine under both MacOSX and Linux. If anyone is willing to put up binaries, or make a port to Windows, that's fine and would be welcomed. The code is under a do-whatever-you-want-with-it license.

Any feedback, and most certainly improvements, are highly welcomed!

Post by **cybereality** » Mon May 03, 2010 4:31 pm

Ok cool. Nice work. I have pretty much abandoned the project but I still want to come back to it some days. Its fun stuff. If I get bored maybe I will port the source to Windows.

Post by **Nebra** » Fri May 28, 2010 2:17 am

Just a quick thought ... what if one would manufacture a parallax barrier for each color, R, G & B and overlay them ... wouldn' this give better results ?

EDIT: I've made a little slide show to make my idea more clear. I think you would loose less light and maybe improve the color ghosting ?

Post by **Erant** » Mon May 31, 2010 3:22 am

It's an interesting idea, but it would require even more alignment. Getting a barrier to align is hard enough as it is. Also, you'd need a color laserprinter for that idea to work, and those usually don't go much above 600dpi. I've found that for certain pixel pitches, 1200DPI is a must (in fact, all the ones I've tried were horrid at anything less than 1200DPI).

It's unfortunate that these barriers are so hard to get right, because they yield very decent autostereoscopic results once you do get them right. I've managed to get some good results out of my barriers anyway. But when looking at your screen right-way-up, the bounding box wherein you get good results without ghosting or color bleed is about 2-3cm wide, and maybe 15cm deep.

I have a cheap chinese tablet coming in, with an 800x480 screen and I'll be trying the barrier technique on that one as well. Not sure when I'll get around to it, as I've also just purchased some Elsa Revelator glasses to fool around with.

Post by **dukenukesam** » Sat Jun 19, 2010 1:21 am

Hi guys, I thought I would give this a try myself and since I dont have a quality printer I thought (could this be tried using a transparent photoshop gif with the barrier printed?). I Used a interlaced picture someone put up here, loaded the barrier in a second layer and something is going on but not wuite 3D. I can see two positions for the Image basically if I move the barrier sideways, but can only see one position at a time. Am I dong something wrong? I read at the beginning of the post that it is "Opaque" my lines are all black, so do I just need to give the lines a bit more transparency to see both positions at the same time?

Post by **cybereality** » Sat Jun 19, 2010 7:36 pm

You need a good quality printer, at least 1200dpi and preferably laser, in order to do this.