3D is not without problems

[icester]

Hello Everybody.

Mathew, do you have a ray diagram of what you are talking about here?
There perhaps seems to be a little confusion about what is being discussed, and I don’t think we/everybody are all on the same page here.

Cheers,

Eric

It is the positive parallax limit@human.stereo.vision.geometry.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

 

[Eric Lange]

I think I agree with a little of what everybody is saying here. Both Leonard and Mathew on different points.

5 degrees angular screen distance is a bit much for me, I can do 4, at five I cannot fuse. At ten meters viewing distance, the two corresponding stereo points can be placed at 870mm apart on the screen (just as Mathew says); in my case closer to 700 mm. I think people forget that eyes/lenses are angular devices and focal length of the eye is a factor as well as the angles of convergence by which eyes can be expected to swivel by in their sockets (to fix their gaze upon a single point). I.e. Objects or specified widths appear smaller the further away you place them…

The scene from Father Ted when he explains to Dougal about cows and perspective.

Not posted on RED stereo forum for the first time (I might add)!

www.youtube.com/watch?v=98hO97ky-sA

I like Leonard’s point in that if you can make it work on the big screen, then it is probably going to work for the small screen, but not necessarily the other way round. The reason for this, ( I believe) is that when you work stereoscopically on a monitor you sit nearly two feet away from it. That’s the angular equivalent of almost sitting behind the back row in a cinema. The front row could be like putting your face six inches away from the monitor. Therefore if it works for the front row in a cinema then it’s probably going to work SAFELY on a monitor or home display, even though it may err on the side of weaker or less dramatic/boring looking stereo. ). I get the point about the stereo “stretching” out to meet you.

I think people need to be more mindful of the total range of parallax (angular or otherwise) that is in a given scene that we can be expected to “ingest” without fusion problems and conflicts/pain and headaches. It’s all very well to talk about BG parallaxes, but what you put “infront” of that is just as responsible if not more so for the headache you are going to get.


I don’t know how many minutes of sustained AGULAR divergence people can take and to what degree (pun intended)?

Cheers,

Eric

 

[Leonard Coster]

Wrong again..
The "65 mm 5 deg projected at 10 m is 873.2 mm"
is always the same regardless of screen size.
Also if you are referring to all wrong projection geometries as styles then you shopuld call it pseudo-S3d styles.
For a human viewer sitting in font of a screen of given distance and size there is only one stereo geometry which must deliver realistic viewer experience of the real world.

Mathew Orman

Hi icester,

Sorry I think you're missing my point.
We are not necessarily able, or even trying to replicate exact natural vision in the cinema.
Often you use different focal lengths, frame rates, filters, all sorts of techniques. There are technical and subjective reasons for this.

My point is that I still want to deliver images that are within the normal ocular accommodation range of the viewer.

You will not necessarily produce a stereo illusion that matches actually being at the film location for the viewer but you will still be delivering a stereoscopic image.

To me, this means keeping the stereo coherent so that your audience is immersed ie not breaking the illusion - which makes the effect less, not more - and not causing brain strain so they can't enjoy the show or have to leave.

I understand the distinction between an accurate stereo reproduction of an environment and one where the stereo geometry is not strictly real world accurate.

Given the many other objectives or limitations on set I would argue that it is not always possible to present a strictly accurate image without making those images very difficult to accommodate (fuse) for the viewer.
In those circumstances, or for other subjective reasons, we may choose to deviate from strict dimensional accuracy of reproduction.

Otherwise we should also ban all lenses that are not standard FoV - also ban off speed shooting etc etc.

All my discussions of background & overall divergence limits are aimed at keeping the stereo image within the range that can be easily accommodated by the audience.

If your aim is to accurately reproduce the dimensionality of the location then this becomes the driving factor in your setups. It will determine your choice of focal length and IOD.
This will limit the range of distances from camera for your subjects and/or where you place the convergence point - but you can achieve it.

I hope I've explained myself clearly - please describe your objectives in case I have misunderstood them.
Whatever they are I'm sure they are achievable - as with everything there is often a compromise but we're used to that in film making

 

[Eric Lange]

Following on a bit from what Leonard said; there does seem to be this “belief” that for good stereo you MUST replicate exactly the human IOD (inter ocular distance), but I have to say, if you do this you end up very with un-engaging stereo for many different types of scenes and circumstances. The Victorians did this with their stereoscopes, and cardboarding and lack of 3d complexity in fine detail was the norm. In fields such as Stereo Photogrammetry (terrestrial and Ariel), good photogrammetrists have been able to push things to the max to squeeze out every last drop of 3d texture out of a recorded subject through the use of very low distortion wide angle lenses, large formats and surprisingly wide base separations (or IODs,( which think is a misnomer)).With experience and knowledge it is possible to create absolutely spell binding stereograms. Distortion correction and judicious alignment and re-sampling can go a long way to create detailed and engaging stereograms, WITHOUT divergence, or Y parallax from distortion, or ranges of parallax and nonsensical horizontal image translations. From a simple point of view one might say that we aim to capture and generate for the viewer the richest 3d percept we can, without carelessly pushing the limits too far (just like what Leonard is saying).

Reconstruction of natural geometry does not allow you to do this. Just in the same way we shoot 2d movies we don’t aim to replicate the same visual experience that a human would have (same format and focal length and angle of vision); if you did this for a whole movie, it would be interesting for a few minutes, but largely un-artful and ridiculous. The notion of replicating “natural” geometry is even more pointless in the case of stereo as it is just as artificial as a 2d movie, by virtue of the fact that it is all projected onto a flat screen. A biological binocular convergent system (the nuts and bolts of human stereopsis) cannot perform “naturally” in this environment; no way no how.

The flat screen really changes the game from nature. If you hold out your thumb you can converge and fix on that point. Objects in the distance are stereo un-fused as blurry double images, but never the less the brain is aware roughly what relative distance these unresolved blurry double images are. If you place your other hand six inches away from your face and fixate still on the thumb at arm’s length the blurry double image of the hand six inches from your face does not conflict with the fused images of your thumb. You cannot do this in a cinematic environment because everything is composed of sharp imagery presented at a single distance on a flat plane. Therefore ALL of the imagery presented on screen has to be SIMULATANOUSLY stereo resolvable/fusable by the human visual system. Therefore the range of parallax presented on screen as compared to nature MUST be very much smaller. It is (I belive) the stereographer’s art to eek out the maximum effect from this narrower range of parallax while staying within comfortable and safe viewing limits. However, I also believe that the human brain is not used to “ingesting” large parallel volumes of broad field stereo fusable data, and this also results in basic “brain tiredness” beyond simple psychovisual mismatches, as in nature you would never normally process the same gigantic volumes of stereo fusable data.

I have to say YES pay attention to safe and comfortable viewing parameters, but “natural geometry” is ALMOST meaningless when set against the artificial constraint of a large flat screen.

Ta

Eric

PS. I don’t think anyone seriously trying to argue that stereo ray geometries that are not “natural” should be called “pseudo –S3d” is going to get very far with that idea and terminology, as “natural” geometry does not really exist when you have a flat screen; unless of course you are displaying a stereogram of a coplanar flat screen (on a flat screen)! It also means that 99.999% of stereo cinematic shots would have to be called Pseudo S-3d, which frankly is not very practical. “Toy Story 3, now in Disney Pseudo 3d, or Psuedo-not quite Real-D”…

In photogrammetry a “ Pseudo-stereogram” is one where the left and right images are swapped to produce an inverted Z of the scene. I think Icester may be referring to that in the context of a 65 mm on screen BG parallax?

 

[icester]

Hi icester,

Sorry I think you're missing my point.
We are not necessarily able, or even trying to replicate exact natural vision in the cinema.
Often you use different focal lengths, frame rates, filters, all sorts of techniques. There are technical and subjective reasons for this.

My point is that I still want to deliver images that are within the normal ocular accommodation range of the viewer.

You will not necessarily produce a stereo illusion that matches actually being at the film location for the viewer but you will still be delivering a stereoscopic image.

To me, this means keeping the stereo coherent so that your audience is immersed ie not breaking the illusion - which makes the effect less, not more - and not causing brain strain so they can't enjoy the show or have to leave.

I understand the distinction between an accurate stereo reproduction of an environment and one where the stereo geometry is not strictly real world accurate.

Given the many other objectives or limitations on set I would argue that it is not always possible to present a strictly accurate image without making those images very difficult to accommodate (fuse) for the viewer.
In those circumstances, or for other subjective reasons, we may choose to deviate from strict dimensional accuracy of reproduction.

Otherwise we should also ban all lenses that are not standard FoV - also ban off speed shooting etc etc.

All my discussions of background & overall divergence limits are aimed at keeping the stereo image within the range that can be easily accommodated by the audience.

If your aim is to accurately reproduce the dimensionality of the location then this becomes the driving factor in your setups. It will determine your choice of focal length and IOD.
This will limit the range of distances from camera for your subjects and/or where you place the convergence point - but you can achieve it.

I hope I've explained myself clearly - please describe your objectives in case I have misunderstood them.
Whatever they are I'm sure they are achievable - as with everything there is often a compromise but we're used to that in film making

At current level of achievements in the field of image acquisition and projection there are many technological solution that can deliver virtual world experience with full rage of human stereo-vision accommodation.
In the next few years the S3D technology will prove that
state of the art geometrically correct methods deliver
experience that is so real that it would be hard to believe that it is only virtual.

What you are proposing is showing a chank of the space placed at distance where it does not strain the eyes of the viewers and flattened enough to allow for wide rage of viewer distances regardless of distortions in scale and perspective.
I call that pseudo-stereo and the only problem in S3D is just that: forcing viewers to watch pseudo-stereo will produce bad psychological and physical results regardless of how small and easy it is to fuse it.
Mismatches in FOV, scale, inertia, acceleration, camera dynamics, specular effects, etc. result in tonally confused viewer experience.

Just in AVATAR one character was shown in so many different scales but none with 1:1 and the impression was just a confusion because if you show a character 10 times smaller then his motion inertia indicate as if he was made out of lead. In next scene the same character is 10 times larger so now the impression is as if he was made out of Styrofoam.
Another gross mismatch was re-texturing of CG characters and compositing. The textures all look as stained paper with zero stereoscopic specular effects resulting in dull look just like all 2D to 3D cheep conversion.
Those are just two of to many mismatches that ware placed in AVATAR by it's creators.
Billboards may work well in 3D but is S3D they look just that: plain flat. Human stereo vision is a precise measurement and navigation instrument and cannot be fulled by 2D tricks. Real life specular textures can only be reproduce in S3D in 3d they just all look like color stains
on matte paper.
If bad S3D like AVATAR is worth millions then perfect
S3D will be worth much more.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

 

[icester]

Eric

PS. I don’t think anyone seriously trying to argue that stereo ray geometries that are not “natural” should be called “pseudo –S3d” is going to get very far with that idea and terminology, as “natural” geometry does not really exist when you have a flat screen; unless of course you are displaying a stereogram of a coplanar flat screen (on a flat screen)! It also means that 99.999% of stereo cinematic shots would have to be called Pseudo S-3d, which frankly is not very practical. “Toy Story 3, now in Disney Pseudo 3d, or Psuedo-not quite Real-D”…

In photogrammetry a “ Pseudo-stereogram” is one where the left and right images are swapped to produce an inverted Z of the scene. I think Icester may be referring to that in the context of a 65 mm on screen BG parallax?

Yes, because you would not want to be called pseudo-stereographer.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

 

[Eric Lange]

Yes, because you would not want to be called pseudo-stereographer.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

That's quite funny;

So I guess that would make Lenny Lipton (for example) a Psuedo-Stereographer as well. Given that I design and build very advanced digital stereo photogrammetric workstations and software aswell as very high end VR apps, I would LOVE to be called a “PSUEDO-STEREOGRAPHER”. I think that’s great. Thank you Icester!

May I recommend the book “How to Win Friends and Influence People?” by Dale Carnegie; Amazon has it for $7.99, might be a worthwhile investment.

www.amazon.com/How-Win-Friends-Influence-People/dp/0671723650


Mathew, just as a foot note we use stereo texturing just in the way you are asking for, a critical eye just as you describe is useful, but also realize that people also seek help on the Forum and it certainly pays to listen. The stuff posted by Pedro on the Synch issue with RedOne is absolutely priceless. Many tens of thousands of dollars could be spent to perform the same tests to get the same valuable information. That's why folks are keen to be fairly supportive at every opportunity as there is something you can certainly learn from everybody, whether or not they have their math right or not, it seems in a lot of cases they are at least DOING it right, and anyone making the case for less extreme stereo (in psychovisual terms), and without seeming patronizing, should be commended and encouraged [as in Leonard's case]!

 

[icester]

That's quite funny;

So I guess that would make Lenny Lipton (for example) a Psuedo-Stereographer as well. !

Wow, that is a perfect score.
If you want to know why, ask Michael Starks he will
tell you who or what is Lenny Lipton under the skin.
Or you might want to research what happened to TEK 3D technology by TEKTRONIX.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

 

[Eric Lange]

Wow, that is a perfect score.
If you want to know why, ask Michael Starks he will
tell you who or what is Lenny Lipton under the skin.
Or you might want to research what happened to TEK 3D technology by TEKTRONIX.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

Icester man, you crack me up; (seriously, I am beaming from ear to ear ( Cheshire cat move over)).

How much gasoline do you want to throw on this one? I am based in New Mexico, maybe I can score some U235 from Los Alamos for you? I guess in the interests of professional decorum, one is not at liberty to ask for such details (no matter how interesting), and things start to cross a line that should never appear in print (especially in a post), but feel free to PM me any pertinent information that you FEEL might warrant a little further investigation in the matter (for the blissfully ignorant), that would not legally constitute slander. Nobody is really going out of their way to deliberately create enemies, but from a technical standpoint perhaps, therefore I guess you would also regard Michael Starks as a well meaning pseudo-stereographer as well?

Sounds like someone has a nice collection of axes to grind; perhaps this is akin to stereo-therapy for angry embittered “true” stereographers.

Where do you go from here (I am laughing) though....?

[Sincerely, thanks for the laugh]

Eric

( I should really ACTUALLY do some work I’ll check in later :beer;

 

[icester]

Icester man, you crack me up; (seriously, I am beaming from ear to ear ( Cheshire cat move over)).

How much gasoline do you want to throw on this one? I am based in New Mexico, maybe I can score some U235 from Los Alamos for you? I guess in the interests of professional decorum, one is not at liberty to ask for such details (no matter how interesting), and things start to cross a line that should never appear in print (especially in a post), but feel free to PM me any pertinent information that you FEEL might warrant a little further investigation in the matter (for the blissfully ignorant), that would not legally constitute slander. Nobody is really going out of their way to deliberately create enemies, but from a technical standpoint perhaps, therefore I guess you would also regard Michael Starks as a well meaning pseudo-stereographer as well?

Sounds like someone has a nice collection of axes to grind; perhaps this is akin to stereo-therapy for angry embittered “true” stereographers.

Where do you go from here (I am laughing) though....?

[Sincerely, thanks for the laugh]

Eric

( I should really ACTUALLY do some work I’ll check in later :beer;

Yes, you are right I will email you directly with attached PDFs that Michael Starks had just sent me.

Getting back on REDUSER relevant subject: I am developing FULL HD stereoscopic adapter for a single RED ONE camera with normal stereo base, off-axis lens geometry and stereo-window adjustable from 1m to infinity. Should have a prototype soon and will post samples hopefully in new S3D blue/ray format.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

 

[Leonard Coster]

Thanks for jumping in there Eric... very brave of you

Hey will you be at NAB? I'm giving a talk there on Tuesday afternoon 2:30 post pit.
I'd love to discuss your work also - VR stations have been into Stereo vision for a lot longer than mainstream projection.

Maybe see you in April.

Cheers,

 

[Leonard Coster]

Icester !

I would love to see some of your footage - soon as you can post any of it please lets have a look.
I have my own 3D monitors and projection so even if you just have left/right eye as eg quicktime I can play that.
Otherwise I can wait for your to present it in whatever forum you prefer.

Cheers,

 

[Eric Lange]

Thanks for jumping in there Eric... very brave of you

Hey will you be at NAB? I'm giving a talk there on Tuesday afternoon 2:30 post pit.
I'd love to discuss your work also - VR stations have been into Stereo vision for a lot longer than mainstream projection.

Maybe see you in April.

Cheers,

I really should stop being lazy and reclusive, and go to NAB, would be cool to see you and your talk, maybe I bring a box to plug in so you can see what all the fuss is about.

Cheers

Eric

 

[Eric Lange]

Yes, you are right I will email you directly with attached PDFs that Michael Starks had just sent me.

Getting back on REDUSER relevant subject: I am developing FULL HD stereoscopic adapter for a single RED ONE camera with normal stereo base, off-axis lens geometry and stereo-window adjustable from 1m to infinity. Should have a prototype soon and will post samples hopefully in new S3D blue/ray format.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

Hi Mathew,

Thanks for the info, I can discuss with you off list, need some time to digest these tomes. Today is my B-day so I’ll have some fun rather than drench myself with the woes of the “3D industry”; Starks seem like an interesting guy and his work is well known in the patent literature.

So perhaps your sentiments and the title of the thread are particularly apt.

“3D IS NOT WITHOUT PROBLEMS!”

Cheers,

Eric

:emote_hippie:

 

[icester]

Hi Mathew,

Thanks for the info, I can discuss with you off list, need some time to digest these tomes. Today is my B-day so I’ll have some fun rather than drench myself with the woes of the “3D industry”; Starks seem like an interesting guy and his work is well known in the patent literature.

So perhaps your sentiments and the title of the thread are particularly apt.

“3D IS NOT WITHOUT PROBLEMS!”

Cheers,

Eric

:emote_hippie:

Happy Birthday!
Hope you shoot some 3D for keeps...

All the best,

Mathew Orman

 

[icester]

Dan,

Read US patent 7643,025 B2, ...

www.uspto.gov/web/patents/patog/week01/OG/html/1350-1/US07643025-20100105.html

to see if you can get it/understand. The application was dumbed down substantially in the hope that the examiner’s might get it. I think then you will get a better idea as to what the user vs the data preparator can sensibly accomplish (excellent question I think you pose), and how “card-boarding” is essentially cured (in the scenario you imagine).

Have fun, the specification is 54 pages long, but I think you might be able to understand/dig it.

Cheers,
Eric

There is a slight problem with that.
The algorithm that does 3d reconstruction from stereo pair will fail on specular reflection, refraction and transparencies. It will only work good on paper textures with ambient lighting.
So then you end-up with some false depth and your intend to reduce parallax by nonlinear or perspective modification of 3D volume will insert new distortion
which ware not present in original stereo pair.
Phillips WOWvx had gone bust because of their attempt to normalize the depth volume to get comforable fixed
range parallax.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

 

[icester]

I think the other thing that causes headaches (which is something that people don’t talk about), is that in nature, when your eyes converge on a point that point is in focus, (for example your thumb at arms length), the back ground will be perceived as a double image, and anything in front of your thumb will also be a double image. The things in the background and foreground are also blurred due to lack of focus (as you are fixing your gaze on your thumb).
Cheers,
Eric

Not quite,
human eye are different than fixed geometry lens.
When you focus on horizon the thumb in front of your nose is out of focus, but when you focus on your thumb
all background including horizon is in perfect focus.
Check it out...
Also, there is nothing wrong with blurred vision, I am 51 and I use +1 diopter glasses for foreground world but I can
take them off and not use it and will never get any headaches or discomfort.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

 

[Eric Lange]

There is a slight problem with that.
The algorithm that does 3d reconstruction from stereo pair will fail on specular reflection, refraction and transparencies. It will only work good on paper textures with ambient lighting.
So then you end-up with some false depth and your intend to reduce parallax by nonlinear or perspective modification of 3D volume will insert new distortion
which ware not present in original stereo pair.
Phillips WOWvx had gone bust because of their attempt to normalize the depth volume to get comforable fixed
range parallax.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

The interface that we have, probably has one of the best manual editing interfaces for stereo immersive feature extraction, and in essence you introduce the more advanced automated algorithms within that context. It has always been my view that such an interface would always live or die by the quality of its manual editing features, simply as right now machine vision is nowhere near as good as the stereo acuity or visual interpretive capability of the human visual system, (but we are working on new automated approaches that are best suited to producing coherently stereo textured models).

The use of stereo textures in the paradigm of coherently stereo textured models in fact embeds specular reflections, refraction and transparencies (as you describe) into the model as it is able to re-present to the respective left and right eyes what you would naturally see (nothing is thrown out!). Therefore a very sparse number of points only needs to be extracted, as a polygonal substrate is used as a framework to support the stereo textures. A judicious selection of these points is required rather than the blanket extraction of many millions of very noisy points. In most instances the polygonal substrate is the basic shape of the object/scene, but the stereo textures are mapped using a network of tri-homologous points. The eye/brain defer to the disposition of the stereo textures in true 3d space once the whole scene is essentially re-rendered as a stereogram. The basic method was originally designed to find a way to accurately record and re-present historic and archeological sites in true 3d/VR systems, (as part of PhD I did at Cambridge University (in the UK) some years ago). The basic problem with historic sites is that you have fantastically rich and complex surfaces over a very wide area that have to be represented in a computationally efficient way and also be metrologically accurate (to less than 1 mm over an entire site). Standard laser scanning and standard photogrammetric 3d automated extraction techniques, and associated data have very real intractable mathematical problems. Therefore I came up with the CSTM which is conservatively 400 times more efficient at representing complex surfaces over conventional explicitly modeled surfaces. It also had the accidental (industrial) byproduct that it allows one to re-render a scene to safer ranges of parallax from a wide IOD WITHOUT compressing the fine 3d textures. The main aspect of the system is designed to join together many stereoscopically captured images to allow one to construct much larger models for example, a whole film set or an ancient cathedral ruin. The interface and software has also been designed to handle multiple camera positions for 360 degree dynamic scene acquisition.

Because the polygonal substrate in true 3d space is essentially invisible to the viewer very few stereo corresponded points are required compared to conventional “auto-extraction” methods. The auto extraction algorithms we are developing are very picky in the sense they have what can best described as a heightened sense of “garbage suppression” as only a small number of points are actually required.

Standard algorithms as you indicate do have real problems, but when you are using CSTMs it’s a new frontier, and new approaches have to be used from the ground up. The technique produces very clean models that support terrific complexity in a computationally very efficient way (in other words things look extremely life like, and you can get away (visually) with murder).

As you can imagine there are many applications.

Thanks for the Question.

Eric

 

[Eric Lange]

Not quite,
human eye are different than fixed geometry lens.
When you focus on horizon the thumb in front of your nose is out of focus, but when you focus on your thumb
all background including horizon is in perfect focus.
Check it out...
Also, there is nothing wrong with blurred vision, I am 51 and I use +1 diopter glasses for foreground world but I can
take them off and not use it and will never get any headaches or discomfort.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/

You must be a lucky man or have really long arms. I certainly can’t do that and I have very good eye sight, I guess everybody is a bit different, or maybe I'm becoming a bit short sighted spending too much time at a computer terminal posting on red 3d forum!.

 

[icester]

The interface that we have, probably has one of the best manual editing interfaces for stereo immersive feature extraction, and in essence you introduce the more advanced automated algorithms within that context. It has always been my view that such an interface would always live or die by the quality of its manual editing features, simply as right now machine vision is nowhere near as good as the stereo acuity or visual interpretive capability of the human visual system, (but we are working on new automated approaches that are best suited to producing coherently stereo textured models).

The use of stereo textures in the paradigm of coherently stereo textured models in fact embeds specular reflections, refraction and transparencies (as you describe) into the model as it is able to re-present to the respective left and right eyes what you would naturally see (nothing is thrown out!). Therefore a very sparse number of points only needs to be extracted, as a polygonal substrate is used as a framework to support the stereo textures. A judicious selection of these points is required rather than the blanket extraction of many millions of very noisy points. In most instances the polygonal substrate is the basic shape of the object/scene, but the stereo textures are mapped using a network of tri-homologous points. The eye/brain defer to the disposition of the stereo textures in true 3d space once the whole scene is essentially re-rendered as a stereogram. The basic method was originally designed to find a way to accurately record and re-present historic and archeological sites in true 3d/VR systems, (as part of PhD I did at Cambridge University (in the UK) some years ago). The basic problem with historic sites is that you have fantastically rich and complex surfaces over a very wide area that have to be represented in a computationally efficient way and also be metrologically accurate. Standard laser scanning and standard photogrammetric 3d automated extraction techniques, and associated data have very real intractable mathematical problems. Therefore I came up with the CSTM which is conservatively 400 times more efficient at representing complex surfaces over conventional explicitly modeled surfaces. It also had the accidental (industrial) byproduct that it allows one to re-render a scene to safer ranges of parallax from a wide IOD WITHOUT compressing the fine 3d textures. The main aspect of the system is designed to join together many stereoscopically captured images to allow one to construct much larger models for example, a whole film set or an ancient cathedral ruin. The interface and software has also been designed to handle multiple camera positions for 360 degree dynamic scene acquisition.

Because the polygonal substrate in true 3d space is essentially invisible to the viewer very few stereo corresponded points are required compared to conventional “auto-extraction” methods. The auto extraction algorithms we are developing are very picky in the sense they have what can best described as a heightened sense of “garbage suppression” as only a small number of points are actually required.

Standard algorithms as you indicate do have real problems, but when you are using CSTMs it’s a new frontier, and new approaches have to be used from the ground up. The technique produces very clean models that support terrific complexity in a computationally very efficient way (in other words things look extremely life like, and you can get away (visually) with murder).

As you can imagine there are many applications.

Thanks for the Question.

Eric

Epilinear geometry supports arbitrary camera position in a scene but still fails on lighting effects regardless of the algorithm used. The ruins are mostly covered with dull specular textures and any 3d reconstruction method woks fine on it. Also such structures
are not in human everyday vision menu so if there are errors and or artifacts the human brain will most likely ignore it or assume as normal shapes of damaged structure.
There is no problem to apply stereoscopic texture on a model, but how do you get correct model from stereo pair in the first step. Also if you do that with CG models then the lighting geometry will conflict with specular effects in stereoscopic textures.
I guess the example with original plus normalized parallax would be a prove that your concept works.

Mathew Orman

For state of the art stereoscopic vision systems with: eyestrain free, distortion free and realistic immersion stereoscopic experience contact me
at:
http://www.tyrell-innovations-usa.com/