Explain shooting parallel plz

[foop]

Im trying to setup a couple point n shoots on a parallel bar so that i can shoot stills. From what ive heard though shooting parallel puts everything in front of the screen while can cause alot of visual conflicts when things touch the sides, not to mention losing out on all the depth behind the screen. Ive read that putting blackbars on the left side of the left view and on the right side of the right view will cause the screen to come forward and push everything behind it, but i tried it on some stereo footage and it didnt do anything.

So what can i do to push everything behind the screen?

[Neil]

Hi Foop!

The solution is actually best handled after the footage is captured.

Here is a guide that will better explain it:

http://www.mtbs3d.com/cgi-bin/newslette ... ews_id=55/" onclick="window.open(this.href);return false;

This was intended for animation capture, but the principles are the same.

Regards,
Neil

[Freke1]

So what can i do to push everything behind the screen?

In StereoPhoto Maker move the 2 images towards the center/dividing line. That will give more depth and eventually no popout. (assuming crosseyed 3D picture).

[foop]

I'm mainly having so much trouble understanding 3d cause every thing i read on it says the opposite of the one before it. For instance the book "3d movie making" says on pg 74:

If your two camera axes are parallel, your 3d picture will be 100% in front of the screen.

which is in direct contrast to the above guide, which shows graphically that shooting parallel will put everything behind the screen. Why is it no one can agree on what happens when you shoot parallel?

You know, now that i think about it the above guide is wrong. It says in the 'depth 3d' section that in a parallel setup the red(left eye) would be on the left and the blue (right eye) would be on the right, but if you look at your environment with only your left eye then everything is shifted right and vice versa, so in the instance of parallel the colors would be switched, correct?

Can anyone with a parallel rig post a photo so i can tell once and for all what the case is?

[Likay]

You can check this thread for some info: http://www.mtbs3d.com/phpBB/viewtopic.php?f=3&t=5261" onclick="window.open(this.href);return false;

Basically: For big screen users with at tiny bit too much convergence the further away objects can't be focused on (the eyes have to converge outwards) while it works with smaller screens.
However: If adapting a convergence suitable for example 50" should suit most people and if bigger screens are used, then a simple horisontal offset could be used in the stereoviewer.

[crim3]

Here is a good guide:
http://local.wasp.uwa.edu.au/~pbourke/m ... reorender/" onclick="window.open(this.href);return false;

[Neil]

Anaglyph is somewhat more complex than what the guide suggests, so color matching won't give you the results you are after.

Parallel shooting does keep everything behind the screen. Proper convergence is adjusted in post-production after you have captured your parallel footage. In stereo photo maker or movie maker, you offset the pictures from each other to determine the convergence point.

Controlling the pop-out has almost NOTHING to do with your cameras. The exception to the rule is if you toe your cameras in, but this causes "keystoning" anomalies, and is not recommended. You may also want to increase your FOV to take into account the ends being cut off during adjustments.

Regards,
Neil

[Likay]

Parallel shooting does keep everything behind the screen.

I hope you mistyped. Parallell shooting keeps everything in front of the screen unadjusted and properly viewed.

[Neil]

That doesn't make sense at all, unless I'm misunderstanding what is meant by parallel.

When I'm thinking parallel, I'm thinking two cameras facing straight ahead, like two human eyes. When I'm talking about behind the screen, I mean depth, like you can reach inside the screen with your hand.

Regards,
Neil

[Likay]

Exactly. When shooting pictures and scenes with the cameras straight parallelled (no toe in or out) the resulting image (when viewed in a stereoviewer) will entirely be in front of the screendepth. Faraway objects will be close to screendepth while closer objects will be even closer to you. This is why horisontal alignment has to be adjusted for more depth later on in stereophotomaker.

It can easily be tested by shooting a scene with parallell cameras and compose the images without realignment in stereophotomaker. Everything will be at prescreendepth when viewed (unless l/r are shifted. )

This is probably the reason why many people means that some amount of convergence should be used when taking pics. Myself i'm a "parallell shooter" but is going to test if any convergence ruins pictures or what imaginally really happens.

Practically (and theoretically because nothing really is infinte): If you imagine taking pictures of a scene from an infinite distance which should suit an infinite big screen for viewing there's only one correct solution for taking pictures and that's straight parallell cameras.

[foop]

From this book i have on 3d production, it says adjusting the cameras interaxial distance in post will only serve to increase or decrease the popout effect, but will never push it behind the screen, and i dont know of any other adjustments you can make in post, so likay, what exactly are you doing to push everything behind the screen?

[Likay]

As for Neil i shoot all my pictures with my cameras parallell. I later adjust the images to both vertical and horisontal correct alignment in stereophotomaker (free excellent software for stereoimages). For my cameras it's about 230 shift in horisontal for indoor and about 170 for longer range images. Keeping at 200 gives a good alignment with proper depth for most images. These figures are for my cameras and differs if other cameras have other resolutions. Test what makes the best result in general. Always keep the original left and right images.
You can easily adjust a bunch of pictures by using the multiconversion function in stereophotomaker so you won't have to realign evry single picture yourself.
Before experimenting with shooting using convergence (at higher convergence you risk ruin your stereoimages) i really recommend shooting with parallell cameras and use later editing in stereophotomaker.

[Neil]

Ahah! Our recommendations are identical, then. Parallel cameras, convergence adjusted after.

Regards,
Neil

[foop]

I assume by 200 you mean pixels (Using the arrow keys right?) so then what resolutions are you working at?

[Likay]

Ahah! Our recommendations are identical, then. Parallel cameras, convergence adjusted after.

Regards,
Neil

Most definitely!

I assume by 200 you mean pixels (Using the arrow keys right?) so then what resolutions are you working at?

The cameras are 10Mpix canon ixus 970is.

[crim3]

A frequent misconception is to think of the cameras as an emulation of the eyes in the moment of taking the shot. What you are really trying to get with the cameras is a representation of what there would be in front of the eyes in the instant of shooting, not emulate the act of viewing with the two cameras as if they were eyes. You are capturing the scene, not viewing with the cameras, visualization will be done with your eyes later. Think that, when the process is finished and you visualize the scene, you can roam around with your eyes to explore the scene at diferent depths as you would do in real life if you were standing in front of the real scene.

The problem arose from the fact that those two images, captured with the two cameras, must be projected in a screen for visualization. As the cameras are displaced when you shot, the cones that emerge from them are also displaced one from other, and so the planes that represents the image plane of the cameras (the plane that cuts the cone at the camera focal point, where the image points lay). But when the two images are shown in a screen simultaneously, these two image planes are forced to lay one over the other without the displacement that existed before, at the moment they were captured.

The result is that points at infinitum will be at screen depth (if the cameras were displaced by viewers interpupil distance) because they are shrunk a distance equal to their original separation and now are coincident, meanwhile the rest will have negative paralax and will appear to be in front of the screen.

The word 'convergence' has some controversy because it may refer to the convergence of the cameras when capturing, convergence of the eyes when viewing and convergence of the images when visualizing, but the three terms are often mixed as if they were the same thing.

(*The guide linked above is not about anaglyph, but generic stereoscopic images generation. It uses the colors to discriminate between left and right avoiding confussions)

EDIT: Assuming there is any thruth in this post, it means that with a dual projector system convergence could be physically adjusted displacing each projector's image. Interesting.

EDIT 2: ...with the side effect benefit of increasing the field of view. Now I understand those professional HMD with two FOV listed, one with full stereo images and another bigger one with non stereo images at the sides. (I don't remember the words in english to describe that... )

EDIT 3: found it: 100% overlap or X% overlap

[foop]

Thats only 5% of your horizontal resolution, not as bad as id thought it would be. Well, i think i got it figured out, thanks.

[foop]

EDIT 2: ...with the side effect benefit of increasing the field of view. Now I understand those professional HMD with two FOV listed, one with full stereo images and another bigger one with non stereo images at the sides. (I don't remember the words in english to describe that... )

I hope your not under the impression that HMD's can give you either stereo or wrap around vision based upon what images you feed them. All you'll get with two halves of a panoramic image on an HMD is a garbled mess.

But i get what you're saying about shooting stereo. It took me some hours to grasp the relationship of the eyes to that of the cameras, and how they are not an replacement but an emulation, so in essence the stereo effect is the mind interpreting the stereo effect of our eyes which are interpreting the stereo effect of the cameras which are interpreting a three dimensional world through a two dimensional medium

[crim3]

I hope your not under the impression that HMD's can give you either stereo or wrap around vision based upon what images you feed them. All you'll get with two halves of a panoramic image on an HMD is a garbled mess.

I mixed different things in the same post, sorry.

My post was a dump of my thoughts after a while infering things, not something that I knew before. Sometimes I dump some trash in the forum in the hope that someone who really knows about the topic corrects me. But all I get is to look like a fool due to my poor english (and is getting worse with time ) and confuse those who try to read it.

After submiting the post I realized that the adjustment made with stereophotomaker, in the case of a dual projector rig, can be done leaving the image with 0 convergence and then moving physically the projectors to adjust the convergence. Later I connected this idea with those expensive HMD that can adjust their optics to select two different field of views: one narrower with a 100% overlap of the two displays and a wider field of view with the two images not overlaping completely. Think that, with high field of views, you don't need stereo at all the image, only at the central zone.