[DIY] Auto-Stereo with Parallax Barriers

[jon08]

Ok! But how to do 3 lines at 16px and then 1 line at 15px? Using the pattern, my whole layer is covered.

Do I need a pattern containing several black lines?


EDIT: I found this combinaison: 5 lines at 16px and then 2 lines at 15px = 15.71. It's ok?
EDIT 2: I found an other combinaison: 12 lines at 16px and then 5 lines at 15px = 15.705882352!! So a difference of 0,0019453609411! It's good or it's better to use small combinaison (because 12 lines + 5 lines, it's bigger than 3 lines + 1 line )

[Fredz]

It won't exactly correspond to your display anyway. First because the pitch may not be exact, second because the glass between the pixels and the surface of the display is refracting the light and altering angles and distances. This depends on the refractive index of the glass and its thickness, and you should also consider the distance between your parallax barrier and the screen. The best solution is to try with something close to what you've measured and to modify the pattern untill it looks ok.



Ça ne correspondra de toute façon pas exactement à ton écran. D'une part parce que le pitch n'est pas forcément exact et d'autre part parce que le verre entre les pixels et la surface de l'écran réfracte la lumière et modifie les distances et les angles. Cette modification dépend de l'indice de réfraction du verre et de son épaisseur, mais tu dois aussi tenir compte de l'écartement de ta barrière par rapport à l'écran. Le mieux est d'essayer avec une valeur proche de celle que tu as calculée et de modifier le motif jusqu'à ce que ça rende bien.

[cybereality]

The pattern needs to be an even number of lines, or it won't tile. Thats why I suggested 3 lines at 16px and 1 at 15px (making 4 lines total). You can experiment with different combinations, but they need to add up to an even number.

[jon08]

Ok. I tried to make a pattern of 12 lines at 16px and 5 lines at 15px but I have a problem. My pattern ends with black which creates a "loop". A picture says more than words:

pattern.png

Ok, I see. I'll try to find a better combinaison

[cybereality]

Read my post above. The pattern needs to have an even number of lines (like 2, 4, 6, 8, etc.).

[jon08]

Yes, I saw yout post later ^^'

I think that 1,75 is the best that we can have (3*16px + 1*15px).

[cybereality]

Only one way to find out...

[jon08]

I go to sleep and I'll try this tomorrow (it's 3:00 am here )

[hlmasterchief93]

@hlmasterchief93: You can watch 3D movies, as long as they can be played with the Stereoscopic Player. So NOT Blu-Ray 3D, but basically anything else.

Im trying to make this for my phone, so no Stereoscopic Player

[cybereality]

Im trying to make this for my phone, so no Stereoscopic Player

You would need to find software that could play 3D videos in vertical interleaved mode. I don't believe there is anything that does this for phones, you would have to create that yourself.

[Fredz]

If you've got an iPhone you can simply buy the iGrilli3D, I don't know it this parallax barrier exists for other mobiles though. For videos you can at least use YouTube 3D which offers a row and column interleaved option. For downloaded videos you'd have to find an application which is able to interleave a side-by-side input or program it yourself as cyber said.

[jon08]

Ok! I have print the barrier this morning. Firstly, my printer doesn't recognize transparent paper So I then hacked by taping a white sheet on the back

And that works!!

So I did some test but the result isn't very good!

Firstly, with the alignement test we can see that my pattern is to big (15.75 but I need 15.70):

IMG_20111108_143346.jpg

Secondly, I have test with some 3D movie and Stereoscopic Player. The result isn't good because I need to move my head too much from left to right (or from right to left), so I search a compromise but that result by ghosting.

What can I do?


EDIT: Can I do a pattern like this: 24*16px+10*15px = 15,7058823529411 (so close to 15,703936992) or a pattern of 34 lines is to big to be precise?

EDIT 2: Ok, I have print with 24*16px and 10*15px. The result with alignement test image show only one rainbow (orange to blue, it's a litte too big again)! I'll do a photo later. Now, I'll test with movies

[jon08]

Ok! This is a photo of the alignement test with my 24*16px and 10*15px pattern:

IMG_20111108_224932.jpg

[Fredz]

You should read this previous message by cybereality in this thread :
http://www.mtbs3d.com/phpBB/viewtopic.p ... 961#p55961" onclick="window.open(this.href);return false;

He explains what to do with your pattern depending on what you see on the screen :


Although you're very close to the calculated value based on the dot pitch, your pattern is too big and must be made smaller. That's what I explained in my previous message, using the dot pitch only serves as a basis for a first approximation. You need to take into account the refraction of the light due to the glass, which has an incidence on the spacing/thickness of the black lines in the pattern.

You could calculate the needed spacing for the barrier, but for that you would need to know the thickness and refractive index of the glass (around 1.5 for glass, 1.000277 for air at STP), which are generally not made public. The best solution is to make several attempts with varying values until you find something acceptable.

[cybereality]

All the math using the dot pitch is only getting you into the ballpark. You are really close now, you just have to experiment. As the guide shows, the barrier is too big right now. Start simple, doing a pattern with 34 lines is probably overkill (and will result in the banding you see in the photo).

I'd suggest starting with a pattern of 4*16px + 2*15px. And then tweaking it depending if its too big or small.

[hlmasterchief93]

Im trying to make this for my phone, so no Stereoscopic Player

You would need to find software that could play 3D videos in vertical interleaved mode. I don't believe there is anything that does this for phones, you would have to create that yourself.

With this setup, the horizontal resolution with be 1/2 ? LCD 1920x1200 will become 960x1200 ?
If i need to make the convert program, i will look like that picture ?

[cybereality]

You can use the Stereoscopic Player in column interlaced mode, or also the IZ3D driver to play games (vertical interlaced).

[hlmasterchief93]

You can use the Stereoscopic Player in column interlaced mode, or also the IZ3D driver to play games (vertical interlaced).

Thanks but as I've said, im trying to make it for phone
Will try and report later

[cybereality]

Well the best thing to do would be do make some pictures using Photoshop (or GIMP) using layer masks with the interlaced pattern. That way you can at least start testing the parallax barrier right away. The images need to be the exact pixel dimensions of the phone, but you cannot be certain the photo app will not apply some sort of smoothing (that will ruin the effect). If you are developing your own program, you can get it to display an image pretty easy I guess. What platform is it you are targeting (iPhone, Android, etc.)?

[jon08]

I think that 4*16px + 2*15px will be too small. How tweak that?

[cybereality]

Well try it and find out.

[nicooke]

I'm quite interested in this method on a bit of a grand scale - 42inch to be exact.

I also thought perhaps anyone who has tried this can contribute their calculations and we can start building a DB - we could then write a simple program to generate a printable image for a given combination and make this easy for the masses

[cybereality]

The problem with large sized screens (ie 42") is that there is usually a fixed distance from the monitor you need to be. This distance can vary, but is around 10-14", which can be too close for a big screen. But maybe with some ingenuity it could be done.

[nicooke]

Ah I see - I'll probably only do this with my PC monitor then - perhaps lenticular may be something to investigate for my TV.

[Okta]

Cyber, have you given any more thought to trying your barrier method on a 5-7 inch display for a HMD? If it works it would out perform the PR4 by magnitudes. Can you get your barrier setup and test with a fresnal lens or even 2 in layers to see how it effects the view distance? If the Fresnel allow closer viewing distance without braking the 3d view it should be a go.
Then you could make a sheet for a cheap ebay 7 inch 800x480, if it works you could use a higher res screen like in Palmers PR4.
Cheers.

[cybereality]

Thats an interesting idea Okta. I certainly would like to do some tests.

[inveni0]

I'm developing an online program to automatically generate a parallax barrier pattern based on your monitor's settings. It saves the trouble of trial and error on getting the lines just right. The good news is that it works MARVELOUSLY (with averaging accuracy down to 1/100kth of a pixel). The bad news is that I can't get a 3D effect. I was wondering how I could get permission to view the attachments in this thread to help me align the barrier. I can see the individual pixels in my screen, and I can see that they line up perfectly (where the eye can't tell) with the individual pixels, but the 3D effect is eluding me.

One thing that may be causing me trouble is that my printer is out of ink, so some lines are blue or pinkish... Could that affect my results?


EDIT: I'm wrong. It DOES work, just not very well. The Parallax Barrier is excellent, but the color of the ink MUST be as black as possible, or the pixels bleed right through the barrier, and that's why my 3D effect is so poorly noticeable. I will finish up the auto PB program and share it here.

[cybereality]

You should be able to see attachments as long as you have an account and are signed in.

In terms of the barrier, yeah, you need black ink. I would also go into the printer properties and make sure only black ink is selected and you use the highest quality with no ink-saver modes (gets the blackest lines). Please take some images of your setup or share the program you made. Thanks.

[Ghostman84]

Any word on that Parallax Calculator program inveni0?

I understand everything about the process except how to make the Barriers.
I have Photoshop CS5 ,but have yet to find any easy to understand instructions to use with it.
If anyone can point in the direction of a really good tutorial that would be amazing.

Also...why is there so much "trial and error" involved? I'm planning on printing out my sheets at Kinko's. Should I just wait for this calculator program to avoid having to make more than one trip?

[inveni0]

Any word on that Parallax Calculator program inveni0?

I understand everything about the process except how to make the Barriers.
I have Photoshop CS5 ,but have yet to find any easy to understand instructions to use with it.
If anyone can point in the direction of a really good tutorial that would be amazing.

Also...why is there so much "trial and error" involved? I'm planning on printing out my sheets at Kinko's. Should I just wait for this calculator program to avoid having to make more than one trip?

Yes, the program seems to work fine, but I have to fix my printer before I can verify that it is as accurate as possible. It should save all of this trial and error, as long as you have the required specs of the screen you're using. But it doesn't generate the entire image... It only generates the pattern. So you'll still have to take care of setting up the document. You just don't have to set up the lines yourself.

I haven't forgotten about it, I just want to make sure that I have it optimized (which I should have finished in the next few days).

[Ghostman84]

Thanks for the reply inveni0.
Is they're anyone out there that can maybe walk me through how to make a barrier in Photoshop CS5?

[Fredz]

There can't really be another solution than trial and error. To print a parallax barrier that is an exact fit for your screen given a specific viewing distance you need to know the exact thickness of the glass and its refractive index. But the only value that is available from the manufacturers is the dot pitch (and often incorrectly even), the other values are never advertised.

Illustration :

[Ghostman84]

Well I guess I'm buying some black ink then! Can someone please give me a nice photoshop breakdown....please!!!!

[cybereality]

There is a tutorial here: http://www.mtbs3d.com/phpBB/viewtopic.p ... 961#p55961" onclick="window.open(this.href);return false;

I know its a little confusing, but maybe you can make some sense of it. I hope to have an easier method at some point but I'm not sure when that will be ready.

[xblack0]

Going to implement the ideas in 3d Editors it will give right answer why some time picture going be mixed


This Project of imaginary 3d world with parallax barriers how (left eye - right eye)

http://www.mediafire.com/?c284shpk8c3p9c7
just use http://www.anim8or.com/main/index.html " anima8tor" to open the file







Result :
1 ) all barriers were same width .
2 ) the closer barrier to each eye center position should be exact above the other image or pixel should be hidden
3) Creating The Barrier shifting method its the distance should each barrier shifted left or right from the
each eye center position which is 90 degree .

http://www.youtube.com/watch?v=MA8a1gV3 ... e=youtu.be

Just i liked to share knowledge maybe i be wrong

thanx

ahmed Turki
United Arab Emirates

cheers

[cybereality]

Very cool.

[twin78]

Great job here i wonder how to make a aligment picture for 1366 x 768 coz my monitor is an acer g195hqv and only supports this
this awesome calculator gives my correct monitor width and height
here

[cybereality]

Hey twin78! Welcome to the forum.

Have you tried my tutorial already? I know its a little complicated, but its worth a shot.

[twin78]

Yeah Cyber reality i tried it you explain real good but its not enough to make an alignment picture written like left right for (oops not 1366 its 1360x768
Thank you for reply

i was able to do one like this: would it work?

[ddadovic]

Hello, I have 1366x768 res.(0.252mm dot pitch). I already made parallax barrier (5.95ppb, but with 12px) with 600dpi printer (i can't afford better right know) and there is very close 3D sense (there is depth, but ghosting when two images are more separated). My brother gave me some clue how to find right interpolation:

firstly i chose 9 lines with 12px and 1 with 10px. 12x9 + 10=118, and on the other side 5.95 x 2 x 10=119. These are not equel, so i must calculate more. And there is formula (x is unknown value here): 118 + 6x = 119 + 5.95x ; at the end x = 20 , so 118 + 120 = 119 + 119
After this, we get match: there must be 20 lines (118 and 119 is already 10 lines on each case, and plus 10 lines we get from the formula (because 20/2=10)), 19 for 12px and 1 for 10px) because with this calculation there is 238 on each side! Can 19 lines 12px and 1 line 10px fool photoshop to reduce 6 to 5.95ppb?

So is this right direction or big misunderstanding?