[DIY] Auto-Stereo with Parallax Barriers

Post by **cybereality** » Tue Mar 06, 2012 8:54 pm

@twin78: That test image looks OK to me. You probably just need to work on the barrier further.

@ddadovic: If 9 x 12px + 1 x 10px is close, then just further adjust it. For example if you need a larger barrier then times the pattern by two and just ad one pixel. So that would be 18 x 12px + 1 x 10px + 1 x 11px. You can do the same thing but make it one pixel less. Hopefully that will help.

Post by **ddadovic** » Wed Mar 07, 2012 10:03 am

Thank you very much!
I totaly skipped some important details from previous pages!

Post by **saldy23** » Sat Mar 10, 2012 10:16 am

Hey! I've been trying to replicate the Parallax Barrier tutorial but haven't met with much success after more than a month of tweaking. My screen's resolution is 1366 x 768 too, with a dot pitch of 0.2512mm. The required barrier width would be around 5.935 pixels. Using a 600 dpi printer, I tried a 6 px barrier followed by a 5px black - 7 px gap barrier. Crashed and burned both times...

Referring to the tips on messing with the odd barrier line here and there to average things out-- by my calculations, I could do 14 lines of 6 px with 1 line of 5px; totals to 89, which is close to 5.935 x 15 = 89.025. Or I'd need 29 lines of width 6px and 1 line of 4 px to get a total width of 178, close to 5.935 x 30 = 178.05 px

@Cyber: I noticed you said that when calculating the interpolation required, the total number of lines needs to be even. That's why I have the two options above. Why's that though?

@ddadovic: Have you had any luck with the interpolation method?

Post by **ddadovic** » Sat Mar 10, 2012 3:32 pm

@saldy23: Well, after printing in local copy store I think I got series of 9 lines of 6px and 1 of 5px on my sheet literally, without fooling the printer to change 6px to 5.95px, probably because I saved file to PDF format (not printing directly from photoshop), and then go to print! When I put barrier sheet to screen, on every ten lines first and final are bolded or something like that, but looking closely everyone is same.

Now, I must wait to get new sheets (inkjet's, to get rid of copy store's depending)... Even like that, there is 3D depth but not without constantly shifting of sheet left-right to find perfect...

Post by **cybereality** » Sat Mar 10, 2012 6:44 pm

@saldy23: The pattern has to have an even number of lines or it won't tile correctly.

Post by **saldy23** » Sun Mar 11, 2012 9:11 am

Thanks Cyber!

But I'm confused about what you said, Ddadovic. Someone please correct me if I'm wrong but isn't the interpolation method meant to fool the human visual system and not Photoshop? I just thought it was more likely that our eyes would not be able to perceive the minute difference in line widths and resolve it as lines of some average width.

Either way, sounds like you're having way better luck than I am

Please may I know what software you guys use to create the base image? I just use this naive method where I take a stereoscopic pair and interleave it column by column to get something like the image attached.

Post by **cybereality** » Sun Mar 11, 2012 12:05 pm

@saldy23: The point of the interpolation is to produce lines that are more precise than the printer resolution can handle. In order to print a barrier using whole pixel values and no interpolation you would need a printer that could do around 15,000 dpi or something crazy like that (which obviously isn't available). Although Photoshop and other programs can do the interpolation for you, I found the results were not as good as doing it manually. However its possible there is an easier method, I still have to test some things.

Also, that image looks OK for a test, though its probably better to use the black/white left/right image as has been posted. This is because that picture of yours is mostly blue (with very little red component) which means it will be difficult to test sub-pixel mis-alignments. But the format looks OK.

Post by **saldy23** » Sun Mar 11, 2012 11:23 pm

I did try your Alignment Test Image, but the closest I got was this...

So I just wanted to know if I would be able to achieve some sort of depth perception even with this sorta close barrier, but I'm not seeing anything with the test interleaved image I showed you in my last post. Was wondering if the base image wasn't good enough as well...

PS The vertical bands you see in the picture are not apparent when I'm viewing it normally, the lines are quite even. Must be my camera

Post by **saldy23** » Sun Mar 11, 2012 11:27 pm

Sorry, putting up a slightly better picture

Post by **cybereality** » Mon Mar 12, 2012 7:50 pm

Yeah, thats really far off. You need to adjust the pattern more, and it may not even be possible with your printer. Whats the DPI on that printer?

Post by **saldy23** » Mon Mar 12, 2012 11:29 pm

I've been using a 600 dpi library printer. I send the command with all the parameters set to achieve maximum quality but I guess the library systems override all this to print in Economy mode. I've tried scaling everything up 4 times to make thick barriers but that doesn't quite work out either. Guess I need to hunt for a better printer huh?

Post by **cybereality** » Tue Mar 13, 2012 8:24 pm

Pretty much. I had similar results with my old 600dpi printer (slightly better, but still).

Post by **Xain** » Tue Apr 10, 2012 12:40 pm

Hi ppl,

i'm a newbie, sorry if i ask but i can't go forward, what you mean with:

"5) Select a 16px x 1px area on the left side of the image." ?

i don't have very much to select, i see only a small line in the middle.

Can someone explain me this point a bit easier?

Thx a lot

Post by **cybereality** » Tue Apr 10, 2012 8:34 pm

You may have to zoom in to select the area. Also, disable "snap to guides", "snap to edges" and all the snapping stuff. You should be able to select an area that is 16 pixels wide and 1 pixel high. Make sure the canvas you are working with is the correct dimensions. I used 32 pixels by 1 pixel in the example, but yours my be slightly different based on your screen. Hope that helps.

Post by **Xain** » Wed Apr 11, 2012 4:47 am

Thank you

i've solved the problem with zooming the image

again thank you.

Post by **Darkultra** » Sun Apr 29, 2012 5:12 pm

Will this work in AMD 3D DLP Checkerboard, or only Vertical Interleaved?

Post by **cybereality** » Sun Apr 29, 2012 6:05 pm

Darkultra wrote:Will this work in AMD 3D DLP Checkerboard, or only Vertical Interleaved?

No, this is just for vertical interleaved. I did try some experiments with checkerboard and it didn't work too well (too much ghosting).

Post by **Darkultra** » Sun Apr 29, 2012 6:40 pm

Thank for reply.

I want to make a Parallax Barrier for my dell inspiron 1420, 1440x900 resolution, screen 14.1, and have a EPSON WorkForce 520 printer. I have an estimated ppb of 11.9565354, and using Photoshop to do 7 lines of 12px and then 1 line of 11px to get 11.875. But If I add one more pixel to the 11px line and made that 1px 50% black less to get a 50% gray pixel, will that work to something like 11.95 or 11.975?

Post by **cybereality** » Sun Apr 29, 2012 7:58 pm

Hard to say. Each monitor is slightly different, so I would say just experiment with it.

Post by **dhirarjanger** » Fri May 04, 2012 11:21 am

cybereality wrote:Glasses-Free 3D Gaming for $5The first thing you need to do is find the dot pitch of your monitor. This will be in millimetres. For example, the monitor I used for this is the Zalman Trimon ZM-M220W, which has a dot pitch of 0.282 mm. Yes, that is actually a 3D monitor, but I was *NOT* using the 3D functionality for this project. So what we need to do is convert this dot pitch into a pixel-per-barrier value. Since printers use dpi (dots-per-inch) we need to convert millimetres into inches. This is pretty simple to do. We just need to figure out how width a pixel is in inches. You can simply type this into Google:

"0.282 mm in inches" [without quotes]

Which Google tells me equals: 0.0111023622 inches

You can also do the conversion yourself, by multipling the dot pitch by 0.0393700787. That will give you the width of each pixel in inches. Now to get the pixels-per-barrier we need to multiply that by the dots-per-inch of the printer. I am going to use the value of 1440dpi, since that is what my Epson does.

dot-pitch(in inches) x dpi = ppb
0.0111023622 x 1440 = 15.987401568

i i am newbie to this but i want to point out that instead of using close values like 15.98 then rounding up why cant i use actual ppi settings i mean my asus moniter manufacturar lists my moniter having pixel density of 96 dpi (same is shown in passmark test) & given that i have printer with dpi of 1440 dpi that gives me 1440/96 = 15 ppb. what i want to point out is that instead of changing units now & then & rounding off the no.'s wouldn't it be great to use ppi & dpi units as these are standerd units for disply.

Post by **Cotelio** » Fri May 04, 2012 4:58 pm

dhirarjanger wrote:
cybereality wrote:Glasses-Free 3D Gaming for $5The first thing you need to do is find the dot pitch of your monitor. This will be in millimetres. For example, the monitor I used for this is the Zalman Trimon ZM-M220W, which has a dot pitch of 0.282 mm. Yes, that is actually a 3D monitor, but I was *NOT* using the 3D functionality for this project. So what we need to do is convert this dot pitch into a pixel-per-barrier value. Since printers use dpi (dots-per-inch) we need to convert millimetres into inches. This is pretty simple to do. We just need to figure out how width a pixel is in inches. You can simply type this into Google:

"0.282 mm in inches" [without quotes]

Which Google tells me equals: 0.0111023622 inches

You can also do the conversion yourself, by multiplying the dot pitch by 0.0393700787. That will give you the width of each pixel in inches. Now to get the pixels-per-barrier we need to multiply that by the dots-per-inch of the printer. I am going to use the value of 1440dpi, since that is what my Epson does.

dot-pitch(in inches) x dpi = ppb
0.0111023622 x 1440 = 15.987401568
i i am newbie to this but i want to point out that instead of using close values like 15.98 then rounding up why cant i use actual ppi settings i mean my asus monitor manufacturer lists my monitor having pixel density of 96 dpi (same is shown in passmark test) & given that i have printer with dpi of 1440 dpi that gives me 1440/96 = 15 ppb. what i want to point out is that instead of changing units now & then & rounding off the no.'s wouldn't it be great to use ppi & dpi units as these are standard units for display.

You use the value that has more significant digits, always use the value with more significant digits. You end up with more accurate results in the end, which will help you get the right interpolation pattern in fewer tries since you're starting off from a closer point. The universe doesn't magically change just because you did some math differently

Post by **cybereality** » Fri May 04, 2012 6:27 pm

Well the thing is, the barrier does not exactly match the width of a pixel. This is due to the fact that there is a piece of glass between the barrier and the display panel. Also you need to take into account your seated distance and possibly the IPD between your eyes in order to get the perfect parallax barrier. So its not just a matter of looking up the reported specs of the monitor.

Post by **djdoom3** » Sun Jun 03, 2012 2:58 pm

why i can not see this attachment? "ZM-M220W_Parallax_Barrier_Pattern.png" ? any idea ?

Post by **cybereality** » Sun Jun 03, 2012 3:06 pm

Its there, its just really small so you can barely see it. Go to this page and just do "save page as" and it should work:

http://www.mtbs3d.com/phpBB/download/file.php?id=792

Post by **djemergency** » Sun Jun 10, 2012 4:12 pm

I finally did it I think on a 22' tv. Does this look ok to you?

[youtube]http://www.youtube.com/watch?v=VWs1vCa6eHI[/youtube]

Post by **cybereality** » Sun Jun 10, 2012 5:25 pm

djemergency wrote:I finally did it I think on a 22' tv. Does this look ok to you?

Yeah, that looks pretty good. Not perfect, but even my mod wasn't perfect.

Have you tried it with any games or videos yet? How is the 3D?

Post by **djemergency** » Sun Jun 10, 2012 5:41 pm

cybereality wrote:
djemergency wrote:I finally did it I think on a 22' tv. Does this look ok to you?
Yeah, that looks pretty good. Not perfect, but even my mod wasn't perfect.

Have you tried it with any games or videos yet? How is the 3D?

Yes it works good. a bit colorful sometimes, but works. There's some ghosting though... mainly in dark scenes where u see lights and stars.

Post by **Cotelio** » Wed Jun 20, 2012 9:48 am

I kind of forgot about this after trying a print once and failing miserably, I was reminded of it when I found the leftover trasnparencies while cleaning, figured I'd try again.

I'm having a bit of confusion with my printer. (DeskJet D1560) It doesn't specify the physical DPI, the only option above 600x600 is maximum DPI, which says black render DPI: 1200 color input DPI: 1200 Color output DPI: 4800x1200-optimized DPI. I can't find any detailed offical documentation on the DPI because HP.

I do have another printer, a Lexmark p4350, ( http://www.amatteroffax.com/xPC_Lexmark_P4350 ) but again, I'm not sure what the physical and optimized DPI is, and with neither I have no idea how to make sure it's printing the max physical DPI without doing that "optimization" BS. I think it's 2400x1200 based on the link in the parentheses, but how do I know if 2400 is the vertical or the horizontal resolution? Another thing is, how do I get past the arbitrary margins and print edge-to-edge?

FYI, monitor I'm doing this too is 1280x768, approxomately 10 1/8 " tall, 14" wide.

Post by **cybereality** » Wed Jun 20, 2012 8:56 pm

Sometimes they consider each color as adding to the resolution (which is why the color dpi will be 4 times the actual black resolution). For that HP it sounds like it may be 1200dpi for black, which should be enough to do this project. The Lexmark also sounds like a 1200dpi resolution. I does 2400dpi max, but it depends which direction it goes in and which direction you need to print the barrier at. For example, if the 2400dpi is the horizontal dpi, and you want to print the barrier vertically, then you can only effectively use 1200dpi. But test it out, it sounds like one of these printers should work.

Post by **Cotelio** » Wed Jun 20, 2012 9:25 pm

I have to play cut&paste with the sheets anyway, I might as well use the highest resolution I have available to me, I just don't know what it is. I suppose I'll just try with my HP for now, I don't actually have a black cartrige in my Lexmark right now, just a photo and a color. I suppose I shouldn't have too much of a problem, the pixels on a monitor just above 720p are a lot bigger targets than the 1980p monitors most people seem to be using.

Still wish someone around here had a wide-format printer though so I could get it with one sheet >: Alas, ( and I believe I said it before ) only place is Staples with a 600 dpi wide-format printer.

I should have picked up that wide-format printer I mentioned before from the garage sale when I had the chance, not sure why I chose to get a pizza instead.

Also, fk. I just put up the test image on the monitor I was going to convert, and it's already all messed up, it looks like like the examples of overscan on Wikipedia. Stupid VGA.... maybe I can get it pixel-perfect somehow first, or else I'm already out on this project, because I can't do it on my main monitor. ;-;

Post by **cybereality** » Thu Jun 21, 2012 5:44 pm

Yeah, it will work best with a digital connection (ie DVI, HDMI). With VGA there is some blurring of pixels that may ruin the effect.

Post by **Cotelio** » Thu Jun 21, 2012 7:48 pm

May be? The absolute best I could get it is looking like an example of over/underscan, for the most part it's solid grey. D: Suppose I can attempt making a barrier for my main monitor, but I won't be able to use it much >:

Worse thing is, only one of my DVI ports actually works for digital video now. Even if I did get a second monitor... stuck with one screen in VGA. Guess I'm overdue for a hardware upgrade.

Post by **dupko** » Wed Jun 27, 2012 4:42 am

djemergency wrote:I finally did it I think on a 22' tv. Does this look ok to you?

[youtube]http://www.youtube.com/watch?v=VWs1vCa6eHI[/youtube]

Nice quality

Post by **Jarvis** » Fri Sep 14, 2012 1:51 am

Seems this threads past tends to resurrect after a period of dormancy..

I've been browsing the forums for a few days and playing around with various barriers from various printers with out "wow 3d" quality success.
To sum it up I've tried the tutorial Cyber posted with various results that seemed to work much better than the java calc app Galo made (no offense but I didn't have much luck so maybe that's my fault haha).

Funny thing about it is the first two patterns I made worked the best.. as in somewhat of a 3d image (very "flat" 3d).
1st was a pattern of 12px with 12px space.
2nd was a pattern of 12px with 12px space then 13px and followed by 12px space.
(this is based off of a laptop monitor at 17.3" and 1440 x 900 resolution)
Anyways now that I'm here, asking for help, and knocking myself in the head, not getting the correct pattern, any idea's?

Onto the experiment:
What formula should be used to factor in most of the variables such as pixel width modification in Photoshop or similar, focal length from observer, distance between actual pixel cell and barrier, etc...?
Does anyone know the pattern for monitor specs ill post under this?

Display specifications for project:
pixels per inch for 17.3" lcd= 98.157
dot pitch=.2587mm (pixels per mm)
width of pixel in inchs= 0.010187760414
ppb= dot-pitch(inchs) x dpi=12.225312497

I have access to some rather expensive canon printers not sure what the dpi is but should be some where upwards of 9600x2400 dpi if not higher and the size sheets that can be printed should be good enough for maybe even a 42" (forgiving size limitations of lcd+parallax combo... also* friend works in a canon showroom with nice $50k+ price tag printing units so should do a fine job once I get the calculations perfected)

Post by **cybereality** » Fri Sep 14, 2012 11:00 pm

Well I don't know if I have any help that hasn't already been said in the thread. Its mostly just a trial-and-error process once you get to a certain point (since each monitor is different). But I wish you the best of luck.

Post by **Jarvis** » Sat Sep 15, 2012 1:09 am

Turns out my screen size was actually 17.1"... store I got it from said 17.3", anyways it worked out because it made finding a pattern much easier.

Results:
pixels per inch for 17.1" lcd= 99.31
dot pitch=0.2558 (pixels per mm)
width of pixel in inches= 0.0100709
ppb= dot-pitch(inches) x dpi= 12.08508 (why the 12px barrier almost worked)

12pixel(11 Lines) + 13PIXEL(1 Line)=132+13=145/12=12.083333333333333333333333333333

12px(22 Lines) + 13PX(2L)=264+26=290/24=12.083333333333333333333333333333
_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_|_
pxpxpxpxpxpxpxpxpxpxpxPXpxpxpxpxpxpxpxpxpxpxpxPX

Other examples:
550 lines and 51 lines = 12.084858569051580698835274542429
*number off by .00023*

12px(183Lines) + 13PX(17Lines)= 2196+221= 2417/200= 12.085
*Number off by 0.00008*

183/200=.915} = 1
17/200=0.085}

100=8.5 +91.5|91.5+8.5=100

Only questionable thing about this is finding a relatively even px(line amount) to work with photoshop...
and I'm gonna give Galo's java app another go around.

Post by **Jarvis** » Sat Sep 15, 2012 1:18 am

I never really post on forums that often just skim through, I've seen a lot of the projects you've done and this one got me excited so what the heck? I'll keep posting to maybe help out other people getting into this hack/mod, very cool stuff!
Btw. How accurate have you needed this to be to work? Curious because this 12px(183Lines) + 13PX(17Lines)= 2196+221= 2417/200= 12.085 is only off by 0.00008* yet it might be a pain to get it worked out in photoshop.. unless you have some tips to set decimal place? or if possible to add and subtract a pixel here or there in the pattern.

If I can get this to work I'll more than likely get a smaller 20-22" 1080p panel and try to position the parallax barrier between the led back light and the lcd screen itself.

Post by **cybereality** » Sat Sep 15, 2012 11:28 am

Well you want there to be some space between the LCD panel and the barrier. The glass is usually enough to do this, without it you won't get a parallax effect. It needs to be in front of the monitor. Because of perspective, the barrier should actually be slightly smaller than the exact dot pitch of the monitor, since it is slightly closer to your face. The distance you sit from the monitor also effects the look of it. Its probably easier to get smaller screens working, though the principle is the same for any size. Any way, hope that helps.

Post by **Jarvis** » Sat Sep 15, 2012 12:45 pm

I figured as much and will try to adjust calculations correctly to suit these parameters. (have some plexy glass and clamping set up to try various distances and size barriers)
On the wiki page for "parallax barrier" they have this:
The newest and most convenient design, commercial products like the Nintendo 3DS, HTC Evo 3D, and LG Optimus 3D do not have the physical parallax barrier in front of the pixels, but behind the pixels and in front of the backlight. They thus send not different images to the two eyes but different light to each. This allows the two channels of light to pass through the pixels, allowing glare over the opposite pixels giving the best image quality.

* Pros
o Clear image
o Largest viewing angle
* Cons
o More expensive for mass production
o Uses 20-25% more backlight than normal displays
Wondered if anyone's tried that with this diy barrier?

Post by **cybereality** » Sat Sep 15, 2012 1:36 pm

Yeah, that would be interesting but I think maybe too difficult (or dangerous) to DIY.