Hugo 3D query

[Gae]

I wonder if anyone can help me with this. I have just written an article on my blog on the use of the stereo window and used Hugo as an example of a film that has a lot of scenes that use negative parallax only. Now on the 3D Blu-ray, it is clear that in several scenes negative parallax was used. An example is in the scene where the theatre audience applauses George Melies. On the 3D bluray, the back of the theatre is in alignment (therefore the stereo window) while everything else is in front in negative parallax. On the YT3D side by side trailer though, (viewing in anaglyph mode) the stereo window is near the front and everything is behind it in positive parallax. Is this just a difference in the anaglyph presentation, or has the parallax been inverted in some way on the 3D Blu-ray?

There must be some 3D geniuses out there who know the answer to this.

Here is my Blog entry.....3D stereo window

Feel free to correct any mistakes in my blog entry.

Gae

[Zivouhr]

I'm not positive, but with anaglyph, the 3D editor can adjust the convergence to compensate for the fact the red/cyan have stronger after images than we'd see with stereo 3D, and so they might have been prompted to "zoom adjust" the convergence slider with the anaglyph version compared to the shot in the stereo 3D version, in an effort to diminish the anaglyph "double image" instead of keeping it the same convergence as the stereo 3D version.

Otherwise, I haven't seen the examples you're referring to, so I'm not positive about this negative parallax situation.

Negative Parallax (3D layer popping out of the screen)
Zero Parallax (3D layer flush with the screen)
Positive Parallax (3D layer deeper into the screen)

[BleedOrange11]

I'm not really a 3D genius. Just speculating...

Unless Scorsese changed the location of the stereo window for the scene after the trailer was released, it's probably just a difference in how YouTube converts SBS material to anaglyph. I'm sure how it looks on BD3D with stereoscopic glasses is how Scorsese meant it to be viewed.

[Gae]

Originally Posted by Zivouhr

I'm not positive, but with anaglyph, the 3D editor can adjust the convergence to compensate for the fact the red/cyan have stronger after images than we'd see with stereo 3D, and so they might have been prompted to "zoom adjust" the convergence slider with the anaglyph version compared to the shot in the stereo 3D version, in an effort to diminish the anaglyph "double image" instead of keeping it the same convergence as the stereo 3D version.

Otherwise, I haven't seen the examples you're referring to, so I'm not positive about this negative parallax situation.

Thanks Zivouhr. The sample I was talking about is the Hugo Blu-ray 3D. Do you have it? If you do, could you have a look at the theatre audience applauding scene and tell me where the stereo window is? On my version I'm pretty sure that the convergence is the back wall of the theatre....I'll have to double check. Even if it isn't though, a lot of Hugo is presented in this way. In some scenes the depth is great and the stereo window is positioned somewhere in the middle of the Z plane (how I prefer it in most scenes) and yet at other times, the scenes seem to be all in negative parallax giving good 3D but less sense of depth.

Originally Posted by BleedOrange11

I'm not really a 3D genius. Just speculating...

Unless Scorsese changed the location of the stereo window for the scene after the trailer was released, it's probably just a difference in how YouTube converts SBS material to anaglyph. I'm sure how it looks on BD3D with stereoscopic glasses is how Scorsese meant it to be viewed.

Yes, I guess the side by side anaglyph method is a totally different creature. It just seemed odd though how the negative/positive parallax balance seemed intact in the YT3D trailer and yet is all negative parallax on the BD3D.

Gae

[BleedOrange11]

Ben Stassen (Sammy's Adventures, Fly Me to the Moon, Haunted Castle, SOS Planet) uses that effect a lot where the stereo window is set at the back of the scene and everything pops way out of the screen. He really pushes it to the extreme, and things get pretty crazy. I like gorgeous depth shots too, but it's pretty neat watching most of the scene take place in your living room.

[Gae]

I'm a bit confused. Can you put the stereo window right on infinity and still use the full 100% of the Z plane depth with just a negative parallax or does it just use a percentage of this, omitting any positive parallax value that might exist? Does the full 100% usage of the Z plane have to include both negative and positive parallax or can either parallax use that 100% on their own, or does it vary? It's important as obviously the stereo depth would be less if only one negative or positive parallax was used as they would only use a percentage of the Z plane depth available.

Gae

[BleedOrange11]

I'm a little confused by your 'full 100% usage of the z-plane' statement.

I think when the stereo window is set at infinity, you get all negative parallax. Everything in the pair of stereo images appears in front of the window and would pop totally out of the screen. Doing this is sort of tricky and is rarely done because the director has to be careful not to create massive window violations.

Maybe I'm all wrong though. Just trying to interpret what I read on this website.
http://www.vmresource.com/camera/stereowindow.htm

[BleedOrange11]

If the director increases the separation between the two images (i.e. increases parallax), the image gets stretched on an anterior-posterior axis.

Wherever the director sets the stereo window is what determines which part of the image will appear in positive parallax (popping out of screen) and which part will appear in negative parallax (behind screen). Setting the stereo window at infinity means everything will appear before the window and consequently pop out of the TV screen.

[Zivouhr]

Originally Posted by BleedOrange11

I'm a little confused by your 'full 100% usage of the z-plane' statement.

I think when the stereo window is set at infinity, you get all negative parallax. Everything in the pair of stereo images appears in front of the window and would pop totally out of the screen. Doing this is sort of tricky and is rarely done because the director has to be careful not to create massive window violations.

Maybe I'm all wrong though. Just trying to interpret what I read on this website.
http://www.vmresource.com/camera/stereowindow.htm

That sounds right.

I have experience with a 3D editor program (Dashwood Stereo Toolbox 3D) that I use with computer animation.

ALL POP OUT:
If you want all negative parallax layers popping out of the screen, then don't touch the convergence slider, and there will be next to zero depth, all at the Zero point but won't reach into the screen (positive layers). And as suggested, the sides of the screen will have major conflicts with one object onscreen while part is not seen at all. Pop out still needs the dual cameras to have enough interaxial distance between each other when filming 3D on the set. For a film like Sammy, the ocean has fewer edges underwater, so they can get away with less convergence adjustment (Convergence being when you blend the two camera images closer or farther away, depending on your goal, usually to eliminate the edge violations where they don't match up and cause visual confusion due to foreground objects too close to the cameras).

DEEP DEPTH:
If you want positive layers reaching deep into the screen, adjust the slider by referencing the closest object to the camera so the two images blend into one. Depth is instantly created even if your two cameras are next to each other, but for great layers, the cameras should be farther apart. You can get 3D Pop into the negative Parallax with parallel cameras provided the foreground object is close enough to the cameras and the cameras are far enough apart, since you're not focusing on a target as you would with converged cameras (which tend to create keystone issues- some object bottoms won't be aligned because of the camera angles) though I haven't tested converged cameras (which I read Avatar used for the 3D rig).

Otherwise, Hugo 3D may have limited how much of the shot they converged and just framed off the edges to avoid window edge issues, but got great pop out (negative layers) at a loss of depth.

Depth, reaching DEEP into the screen, for parallel cameras really benefits by converging the two images with the convergence slider. The more you converge it, the deeper into the screen you can reach, at the cost of less pop out. That's why so many movies stay within the zero zone, not going beyond the screen very often. Convergence is often required to get rid of window violation issues at the edges.

Not sure if that helps. I don't have Hugo 3D yet as it wasn't my style of movie genre but still a well made film, but have seen it in theaters and it's definitely impressive. I'll buy it when it hits 20$ or so.

[Gae]

Originally Posted by Zivouhr

That sounds right.

I have experience with a 3D editor program (Dashwood Stereo Toolbox 3D) that I use with computer animation.

ALL POP OUT:
If you want all negative parallax layers popping out of the screen, then don't touch the convergence slider, and there will be next to zero depth, all at the Zero point but won't reach into the screen (positive layers). And as suggested, the sides of the screen will have major conflicts with one object onscreen while part is not seen at all. Pop out still needs the dual cameras to have enough interaxial distance between each other when filming 3D on the set. For a film like Sammy, the ocean has fewer edges underwater, so they can get away with less convergence adjustment (Convergence being when you blend the two camera images closer or farther away, depending on your goal, usually to eliminate the edge violations where they don't match up and cause visual confusion due to foreground objects too close to the cameras).

DEEP DEPTH:
If you want positive layers reaching deep into the screen, adjust the slider by referencing the closest object to the camera so the two images blend into one. Depth is instantly created even if your two cameras are next to each other, but for great layers, the cameras should be farther apart. You can get 3D Pop into the negative Parallax with parallel cameras provided the foreground object is close enough to the cameras and the cameras are far enough apart, since you're not focusing on a target as you would with converged cameras (which tend to create keystone issues- some object bottoms won't be aligned because of the camera angles) though I haven't tested converged cameras (which I read Avatar used for the 3D rig).

Otherwise, Hugo 3D may have limited how much of the shot they converged and just framed off the edges to avoid window edge issues, but got great pop out (negative layers) at a loss of depth.

Depth, reaching DEEP into the screen, for parallel cameras really benefits by converging the two images with the convergence slider. The more you converge it, the deeper into the screen you can reach, at the cost of less pop out. That's why so many movies stay within the zero zone, not going beyond the screen very often. Convergence is often required to get rid of window violation issues at the edges.

Not sure if that helps. I don't have Hugo 3D yet as it wasn't my style of movie genre but still a well made film, but have seen it in theaters and it's definitely impressive. I'll buy it when it hits 20$ or so.

Wow Zivouhr, you really know your 3D. Could you please read my recent blog entry and tell me if I'm understanding the whole Z plane theory correctly? When I converted Carnival of Souls and part of House on Haunted Hill to 3D I drew hundreds of depthmaps using the greyscale gradient of 0 to 255 to create the 3D effect. Now my idea was, that if you use this greyscale amount and correlate it to the Z plane then 0 should be infinity and 255 would be in your face. So when I'm talking about using 100% of the Z plane that is what I mean where 0 = 0% (infinite positive parallax) and 255 = 100%. Therefore if the stereo window was set exactly at midpoint (127.5 on the greyscale and 50% on the Z plane) then the depth and pop-out would have exactly the same size. If the stereo window is set at the point of say a theatre wall in a 3D movie, does that mean that the negative parallax is only using 50% of the available Z plane amount or is it dependent on other elements like lens separation etc? I'm fascinated by this because I feel that if everything is in front of the stereo window, then you are only using 50% of the available Z plane as our eyes are being deprived of the 50% of the Z plane that is positive parallax which would theoretically be out of our view behind the wall.

Does that make sense?

Here's my Blog entry 3D Stereo Window

Gae

[BleedOrange11]

I'm pretty sure you could stretch those negative parallax-only images starting from the plane of the screen all the way to the viewer's eyes. Obviously, if you set the stereo window in the middle of the scene, you could double the amount of z-plane distance it stretches, but that would look totally unnatural.

Here's Haunted Castle on YouTube if you want to see it in action (uploaded by a Dutch company that seems to own the film now).

Haunted Castle - FULL HD - Real 3D! - YouTube

[Gae]

Originally Posted by BleedOrange11

I'm pretty sure you could stretch those negative parallax-only images starting from the plane of the screen all the way to the viewer's eyes.

Obviously, if you set the stereo window in the middle of the scene, you could double the amount of z-plane distance it stretches, but that would look totally unnatural.

Hmmmm, that's not how I see it if you read my blog entry. I think the most natural type of 3D is exactly that one where the stereo window is in the middle and you have objects on it, behind it and in front of it. To me, that would look natural and wouldn't stretch the images. All you are doing is using both positive and negative parallax at the same time....perfectly OK in the way I'm reading it. Of course, it all depends on the scene you are shooting, the position of the characters/objects and the effect that you want to create.

Gae

I think the confusion lies with the definition of Stereo Window. I guess some people believe it to be the actual screen on a 3D TV or in the Cinema which may be true. My understanding of it is that it's more to do with the actual alignment of the stereo pairs within the screen image and our perception of where it is....so in theory, the stereo window could actually appear slightly behind the actual physical screen depending on the scene and where the convergence is within that image. For example, if two people are standing a couple of feet in front of the camera and are in alignment, then the stereo window will appear to be two feet behind the physical screen. Therefore anyone who walks in front of these two character will automatically appear in negative parallax but still slightly behind the physical screen or am I wrong in that assumption and would that be a window violation? Does the stereo window always have to be positioned on the actual physical screen on a 3D TV or in the Cinema?

Gae

[BleedOrange11]

Sorry, I should have been more clear.

It's not unnatural to set the stereo window in the middle. It's unnatural to have foreground objects extending 100% all the way to the viewer in negative parallax, and then have background objects recess that same magnitude behind the screen into positive parallax. That would be way too much 3D.

[BleedOrange11]

Originally Posted by Gae

Does the stereo window always have to be positioned on the actual physical screen on a 3D TV or in the Cinema?

Gae

Yes. The stereo window defines what is positive and negative parallax. You can't have positive parallax that appears in front of the window.

http://www.dabiri.8m.com/Stereo-Window/

[Gae]

Originally Posted by BleedOrange11

Yes. The stereo window defines what is positive and negative parallax. You can't have positive parallax that appears in front of the window.

http://www.dabiri.8m.com/Stereo-Window/

Yes I know but what I'm trying to say is that the stereo window can be positioned anywhere and not just on the physical screen. Take a look at this screengrab from Madagascar 3 for example. In this instance the stereo window is on the character because he is in alignment (there is no double image) and therefore in the middle of the Z plane. Of course everything in front is now in negative parallax and everything behind is in positive parallax. Therefore the stereo window isn't on the actual physical screen of the TV just where it is positioned within the actual scene and how we perceive it relatively speaking. I guess this has window violations because the foreground objects are cut off. That's quite limiting for object placement I guess. I suppose there is an argument that the perceived stereo window still appears to be the physical screen as that's where we get our reference from......hmmmm...confusing stuff this 3D!



Gae

[BleedOrange11]

Using your example values, +255 is a position behind the screen/window in positive parallax. 0 is the screen. -255 is a position of the same magnitude in front of the screen/window.

Objects at zero (same plane as screen) will perfectly overlap if you put the stereo pairs on top of each other. If a right eye object is placed to the right of the left eye object, the stereo image will appear in front of the screen/window. If a right eye object is placed to the left of a left eye object, the stereo image will appear behind the screen.

[Gae]

Thinking about it...maybe we do always perceive the stereo window as the physical screen too. That would make sense I guess.

No wonder the studios are having so many problems creating great 3D!

Gae

[BleedOrange11]

In the image you posted, the zebra is mostly at the plane of the screen/stereo window. The cake and the boxes and barrels in front of him pop out of my computer screen into negative parallax.

[Gae]

Originally Posted by BleedOrange11

In the image you posted, the zebra is mostly at the plane of the screen/stereo window. The cake and the boxes and barrels in front of him pop out of my computer screen into negative parallax.

Yes, I see that now. So its almost impossible to not have a window violation with negative parallax objects that are in the corners. They will always have to stay central on the screen...that's quite limiting. It probably also explains why most negative parallax usually involves a spear, ball, snake, hand or other object coming at you!

Gae

[Zivouhr]

Although the 3D in Hugo looked very nice and had a pop to it, the overuse of negative parallax meant that it missed out on some 3D depth in many scenes.

-Quoted from http://3dperfection.blogspot.co.uk/p...eo-window.html

Originally Posted by Gae

Wow Zivouhr, you really know your 3D. Could you please read my recent blog entry and tell me if I'm understanding the whole Z plane theory correctly? //I'm fascinated by this because I feel that if everything is in front of the stereo window, then you are only using 50% of the available Z plane as our eyes are being deprived of the 50% of the Z plane that is positive parallax which would theoretically be out of our view behind the wall.

Does that make sense?

Here's my Blog entry 3D Stereo Window

Gae

Great blog to read. Thanks.

I'm not sure if Hugo 3D was filmed with converged cameras that turn inward towards each other, both pointing directly at the target focus object in the shot (more common for live action), or if they used the parallel camera 3D rig on the set.

Either way, when the film is played back, and if no adjustments are made in a 3D editor, it's true that if you had foreground, mid and background objects in view, and your cameras were far enough apart to get strong layers, there's a high likelihood the foreground objects at the sides of the screen will be "messed up" where the head might be in view, but in the other camera eye, it wouldn't even be visible, causing focus confusion, as the eye doesn't know where to focus at the edges. This is mostly the case with parallel cameras, since converged cameras both "look" at the target object.

The way to correct that, and get stronger depth at the same time, is to ZOOM crop while adjusting the convergence slider in a 3D editor, so you merge the two foreground pieces visible into almost one object (which tends to flatten it out, but corrects the problem by cropping out and zooming in), so parts will be cut out completely in an effort to eliminate the visual edge confusion.

On the set, in that situation, to avoid that problem, they'd have to bring the cameras closer together, since the foreground object always determines how far apart you can comfortably set your 3D cameras interaxial distance. The closer the foreground object is to the cameras, the closer the cameras need to be to each other, sacrificing a bigger percentage of depth and layer pop into the screen.

Computer animation, sliding the cameras farther or closer together CAN be done in one shot depending on the foreground's position and visibility in that moment in the shot. Live action, sliding them actively during a shot, say zooming towards a target object, far back, you can have the cameras set far apart for better layer pop, and bring them closer together as you near your target object, so it stays IN VIEW in both cameras.
Had you not adjusted the cameras and just zoomed in, the target object would be visible in the lead camera, but completely out of sight in the other camera eye (for parallel, but not a converged 3D rig, which doesn't need much post 3D editing convergence, but needs "keystoning" corrections). I actually only work with parallel cameras for computer animation stuff.

This is a complicated discussion, but I bet it can help those stereographers new to 3D filmmaking.

By the way, your 2D to 3D conversion of that black and white movie shot looks awesome Gae. Great layer pop, which is number one in my book.

Originally Posted by BleedOrange11

Sorry, I should have been more clear.

It's not unnatural to set the stereo window in the middle. It's unnatural to have foreground objects extending 100% all the way to the viewer in negative parallax, and then have background objects recess that same magnitude behind the screen into positive parallax. That would be way too much 3D.

Great points made in your posts.

Originally Posted by Gae

Thinking about it...maybe we do always perceive the stereo window as the physical screen too. That would make sense I guess.

No wonder the studios are having so many problems creating great 3D!

Gae

Yeah, it can get pretty complicated at first, but once the stereographer gets the hang of the rules, they can get it as they like.

Originally Posted by Gae

Yes, I see that now. So its almost impossible to not have a window violation with negative parallax objects that are in the corners. They will always have to stay central on the screen...that's quite limiting. It probably also explains why most negative parallax usually involves a spear, ball, snake, hand or other object coming at you!

Gae

Yes, center on the screen is key if there isn't going to be post 3D convergence adjustments with the zoom/cropping option. If your foreground object is on the side, then it has to be a bit flatter since the two images have to be combined almost into one object in both eyes, and then everything into the screen looks like a double image with the convergence adjustment (usually for a parallel 3D rig).

3D Rig Converged Cameras, common for live action. Picture a "V" where both cameras turn inward, to focus on the point of "V". Only cost is "keystoning issues", where "X" represents the top or bottom edges of either the focus object or other objects in the shot, so instead of a straight line "I", the two objects mismatch and create an "X". Post 3D editor adjustments can align them properly. Convergence is already mostly accomplished here.

3D Rig Parallel Cameras, common for computer animation. Post 3D convergence is necessary in most cases, to get depth, but also get the foreground edge object to meet and avoid eye confusion if the cameras were too far apart after it was already rendered.