Better 3d to 2d conversion.

[VideoKhahh]

Hello. I am a brand new member so I am not sure how to find out if this question has already been asked. I have a 3d video, "History of the world in two hours" and even though I have found the 2d version (not as high a resolution), the question came about as how to best convert a 3d video into a 2d video.

I have found some answers online dealing with the issue, but none of them preserve all of the resolution of the original file. Most solutions crop the 3d version to create a 2d video. This effectively halves the original file. If the original 3d resolution is 1920x1080 and you crop it, you have a file that is 960x540 that you enlarge to fill the 1920x1080 area.

As I understand it, a 3d movie, even considering the different ways this is accomplished, at 1920x1080 is two alternating frames at 960x540 that when combined is 1920x1080. Is there a way to "mesh" (interweave) the alternating frames to create a full resolution 2d movie?

Now, I understand that there are different formats - Passive vs Active - my file is of the side by side version so If it was possible to interweave the two halves my file would be possible to so. Is this possible? Or am I missing something about how a 3d movie is displayed on screen.

Thanks (this is my first post).

[newpball]

If the original 3d resolution is 1920x1080 and you crop it, you have a file that is 960x540 that you enlarge to fill the 1920x1080 area.

As I understand it, a 3d movie, even considering the different ways this is accomplished, at 1920x1080 is two alternating frames at 960x540 that when combined is 1920x1080.

You may want to revisit that.

By the way the left and right view of the video are different thus merging them will obviously only give a gibberish mess.

[Cornucopia]

You are missing a number of things:

3D has various formats in their capture/creation, in their storage, in their transmission/distribution, and in their display. You are conflating one variety with another.

If your 3D file is stored in SbS layout, it DOES NOT also store it in alternating frames format. It only might DISPLAY it in alternating frames format (what's known as "Active/LCS 3D").
SbS has 2 variations: one that is backwards compatible to standard 2D frame sizes (e.g. 1920x1080, such that each eye view is anamorphically squeezed into their separate 960x1080 halves), and one that isn't backwards compatible (e.g. 3840x1080). Sounds like yours the former.

This means if you want to extract 2D from this (and this is your only copy of the source), you MUST crop to 960x1080 (either the L or the R side - doesn't usually matter) and then resize (anamorphically unsqueeze) to 1920x1080. You might think it isn't full resolution/quality compared to a normal 2D, and you'd be right. Them's the breaks. If you want full rez, you'd need to either get an MVC source, a non-backwards-compatible SbS source (or similar layout), a discreet 2-file/2-stream source, or just get a (better) 2D source.

BTW, you misunderstood how to crop in your original example, it wouldn't have been 960x540.

Scott

[pandy]

Side to correct Cornucopia info i would try approach with different aspect without resizing (unsqueezing). This may save some quality.

[Cornucopia]

Who would want to view a show where the PAR is 1:2? Unless one doesn't care about stick-thin people and "smartcar"-sized elcaminos, it has to get unsqueezed at some point prior to display (and displays currently aren't designed to do that - even the 4:3->16:9 stretch mode isn't enough). This can easily be empirically demonstrated.

@pandy, instead of pointing out "what might theoretically/hypothetically be possible", why not point out what makes practical sense?

Scott

[pandy]

Who would want to view a show where the PAR is 1:2? Unless one doesn't care about stick-thin people and "smartcar"-sized elcaminos, it has to get unsqueezed at some point prior to display (and displays currently aren't designed to do that - even the 4:3->16:9 stretch mode isn't enough). This can easily be empirically demonstrated.

@pandy, instead of pointing out "what might theoretically/hypothetically be possible", why not point out what makes practical sense?

Scott

Any decent SW/HW will support such anamorphic aspect - this is not theory but practice.
That's all from my side on this topic.

[Cornucopia]

Many SW/HWs do support Anamorphics, but you know as well as I and many others here that MANY SW/HW don't. Heck these forums are littered with people asking what to do BECAUSE they can't get that to work.

You would still have to re-encode (since you're cropping), AND you would have to reflag as 2:1 PAR (instead of the existing 1:1) to get it properly to those apps/devices. So quality lost anyway.
PLUS, there are many of those that, while they may support other ARs besides 1:1, they often only support certain specific flavors: 11:10, 9:8, 4:3...
2:1 is not a commonly expected one, so not commonly supported.

Again, you are throwing out tangential hyperbole instead of pragmatically addressing the situation and helping the OP.

Scott

[pandy]

Again, you are throwing out tangential hyperbole instead of pragmatically addressing the situation and helping the OP.

At first i don't see any justification to reduce possibilities (especially without detailed knowledge about OP environment), second i assume OP can perform test and choose what better suit his needs.
Over&out

[jagabo]

Resize with nnedi3. That will give you better upscaling than any HDTV or Blu-ray player. And you won't have to worry about the player being capable of anamorphic resizing.

[VideoKhahh]

Thanks for all of the replies and interest.

I see that I wasnt clear on how the movie was created. I felt that all 3d movies were stored in an alternating frame file and didnt realize that that is'nt always the case. I thought that 3d was somewhat based on the ole ntsc format where video was displayed in alternating frames that merged to produce the final picture. I then didnt make the connection that 3d displays slightly different perspectives of the picture to create the depth that is 3d.

So, if the 3d picture was made up of alternating frames and the slightly different frames were combined you would wind up with a softer not sharper picture because the now interlaced frames were created with slightly different perspectives

It is also now understood that a 3DHD movie isnt really true Hd. I didnt realize that. I was under the impression that since there were two half HD frames that created the full hd picture when processed through the 3d food chain. Not true. A 3D movie is effectively a half Hd movie filled to a full hd screen.

Unfortunately this means that cropping is necessary and you have to throw out perfectly good data. What a waste
Wish there was some way to include it in the final copy....................oh yeah, that's how this thread started

I will try Nnedi3 though. Its just something I want to try, to see what it can do. Just to experiment.

So when 4k 3d comes on the scene there will finally be true 1920x1080 3d........but again at half the displays resolution.

Which raises the question, is there any current HD3d format that displays at full hd?

[pandy]

At first this is not 3D but stereoscopy,
at second not sure why you reduce number of pixels by 4 where reduction is equal 2.
Finally http://en.wikipedia.org/wiki/DVB_3D-TV all format provided - as such there is plenty of formats providing FHD.

[jagabo]

I belive Blu-ray's MVC format is full resolution.

[VideoKhahh]

Thank you for your info, but stereoscopy is where we are at today. Most systems use it.

I was only making the connection about the requirements for full hd using side by side or top and bottom systems in todays current environment. In order to get full 3dhd you have to double the resolution in order to use the current system - more or less. Gotta go.

Thanks though.

[Cornucopia]

Gotta debunk a few things here:

1. Stereoscopy IS 3D, just a 2-view, limited form that shows only a single viewpoint's "projected" recording. Witness that the great majority of 3D's dataspace can be recovered from just those 2 views (with the obvious exception of occlusions).
2. VERY FEW videos/movies are stored in Alternating Frame format. And, what you describe is NOT alternating frame but alternating FIELD format. That was common when functional 3D was restricted to Analog SD (with its interlaced fields). Just as Analog SD is a veritable dinosaur, so it Field Sequential 3D.
3. No forms of Stereo3D "merge"! (except in your brain, where they are projected into recovered true 3D impressions). Part of the difficulty of 3D is in the maintaining of the distinct separateness of those 2 views. Without this strong distinctness, you have "crosstalk" (aka "ghosting") and blurriness. This is one of the main reasons that the Anaglyph stereo3D layout family is such a disappointing "lowest common denominator" format - it has a very hard time maintaining separateness (due to the color filter inaccuracies).
4. If S3D alternating frames ARE combined, you would get a sort of "double image" - NOT blurry in some sections, blurry if some other - that is disconcerting because it is obviously WRONG.
5. There are a number of Stereo3D formats/layouts that ARE true/full HD. Dual-file is. Dual-stream is. Frame-sequential (aka Page-flip) is. MVC is (though with compression). Non-frame-compatible SbS & TaB are. Frame Packing (a quasi-FrameSequential/quasi-TaB format) is. A few other formats are. However, just about all of those are NOT 2D frame compatible, and only MVC (and to a much lesser extent, Dual-stream) is player backward-compatible. This is why the common 2D frame-compatible, player backward-compatible layout/formats (Anamorphic/Half SbS & TaB, and Field Sequential) have remained so popular. And those do lose quality.
6. To clarify: Blu-ray stores their S3D in MVC format, thus retaining ~FULL HD for both views*. It is decoded to individual complete views and then reformatted before being transmitted via HDMI using the Frame Packing layout/format. Then it is displayed either via Active (LC shutter) format, Active+Passive (RealD Z-screen) format, or Passive (FPR Circ. Polarized) format. Of those 3, only the FPR passive HD display does not retain Full HD for both eyes. However, on 4k displays, FPR passive IS Full HD. This is one reason why Blu-ray is important to S3D movies, as it is quite possible to retain Full HD (each/both eyes) all the way through the chain.
7. If you take a (MVC) BD3D rip and convert it to NBC (non-backward-compatible) Full SbS and store it on your hard drive, play it via Stereoscopic Player, etc., and send it out via your heavy-duty video card's HDMI 1.4 port to an Active 3D display (or a 4k Passive 3D display), you WILL see it in Full HD both eyes. Again, squeezed SbS/TaB are compromises for compatibility's sake.

Scott

*MVC uses Main+Dependent views. Dependent view has some inaccuracies, so a very narrow understanding may consider it's view to not be fully the same quality as the Main view.

[Cornucopia]

stupid duplicate!

[VideoKhahh]

Hello again.

Thanks for your detailed info.

As far as some of the updated facts that I suggested/thought were correct, you are correct but for my purposes - someone that asked a question because they are trying to find an answer - the words that I chose were the ones that i thought properly expressed the points that I was trying to explain. The difference between "blurry image" and "double image" would be the same to me if you hadn't made the proper distinction.

I already knew that there are other tru hd formats, the conversational arena was more leaning towards the consumer formats. The other formats arent as well adopted for the consumer.

For all intent and purposes if a 3d movie is stored full hd but because of the delivery chain it becomes less than full hd for both eyes, then the movie as displayed isnt full hd. Thats why I made the statement that only on 4k displays would you get full hd to both eyes. I still wonder if 4k may increase 3d sales because of this feature.

Since you seem to understand a lot about 3d and how it is displayed, could you answer this question that I posted on another thread concerning where the two side half screens (top/bottom - side by side) converge into one full screen image that we see. Here is the last post that I made to try to explain what I had me mystified.

https://forum.videohelp.com/threads/371853-How-do-the-two-main-3d-files-types-manage-to...22#post2390222

Hello again. I read the referenced page and the question still holds. This is the dilemma that I was referring to. Of all of the pages I have read instructing how 3d works not one page answers my question, and I have been reading about 3d since its inception. That to me is what a dilemma is about. Nowhere am I getting the answer that fits.

I am asking how the rubiks cube is made not how to solve the puzzle. The cube has many moveable parts than somehow all stay together. How can such an object be made( actually I know the answer to this)? This is an illustration of what i am asking.

Let me try again.
In a top and bottom 3d video there are two images stacked. Both are 1920x540 in size. There is one image at the top of the screen 1920x540 in size and there is the other image 1920x540 in size at the bottom. I the viewer, with out my magical 3d glasses see the two individual images, the top stacked on the bottom image.

Lets say that the image is of a building that goes from top to bottom, no extra space. The bottom image has a building that that extends from zero to 540. The top image has an image that extends from 541 to 1080

When I put on my 3d glasses the 1920x540 image at the top and the 1920x540 image at the bottom somehow become one 1920x1080 image full screen.

Where does the merging and full screen come from? Where does the magnification of the two separate images and merging come from. It doesnt seem to come from the glasses but thats the only thing that is different from the original viewpoint. If I put the glasses on and look at the world it isnt magnified, yet somehow two buildings 540 pixels in size has become one building 1080 pixels in size from two half size buildings.

If I set a picture of the eiffel tower as wallpaper on a monitor and then take the same picture and set it as wallpaper on a monitor twice the size the wallpaper is half the size of the screen. Yet in the 3d world if I take two pictures of the eiffel tower and set them as wallpaper I get one building twice the size of the original picture. How does this happen?

Oh boy, I knew this would be hard to describe. But please let me know your thoughts. Its still a puzzle to me.

Do you understand my question? Do you have any thoughts?

Thanks for your assistance.

[Cornucopia]

For all intent and purposes if a 3d movie is stored full hd but because of the delivery chain it becomes less than full hd for both eyes, then the movie as displayed isnt full hd. Thats why I made the statement that only on 4k displays would you get full hd to both eyes. I still wonder if 4k may increase 3d sales because of this feature.

This is still incorrect. 3DBD (FullHD in MVC stream)-> HDMI 1.4 (FullHD in FramePacked transmission format) -> Active LCD 3D HDTV display (FullHD in 2xPageFlip/FrameSequential mode). Nothing is 4k, Nothing is non-standard consumer equipment, yet it still is FullHD both eye end-to-end.

...will further respond in a moment...

Scott

[jagabo]

where the two side half screens (top/bottom - side by side) converge into one full screen image that we see

In your brain. Your left eye sees the left 2d image, your right eyes sees the right 2d image, your brain decodes that into a single 3d view.

[Cornucopia]

Wish I had seen this earlier, I would have spared you some of the difficulty of that interaction.

I've mentioned this before:
There are multiple methods for capturing S3D,
multiple methods for editing/storing S3D,
multiple methods for transmitting S3D
and multiple methods for displaying S3D.

RARELY do they "match up" end-to-end, so you should forget trying to categorize them only that way. Also, even though there are multiple options in multiple stages, it doesn't necessarily mean that quality is lost going from stage to stage (just a strong possibility unless you're careful).

Your understanding of S3D movies leads me to assume you are getting the bulk of your S3D sources through the internet/streaming and cable/sat. Those are the distribution channels where Frame-compatible (Half-rez) SbS and TaB are most popular. However, they are NOT the only method (e.g. 3DBD, DVB w/2D+depth, DVB w/2D+delta, or DVB w/3DZ "3DTile", Interlaced, Checkerboard...), CE-centric or not.

What you must understand is that, barring the lucky rare coincidence of the format being identical throughout (Interlaced?), you must always TRANSLATE your layout/format from one stage to the next. This usually comes in the form of temporarily DECODING, RE-FORMATTING/COMPOSITING (usually with 2-view compositing planes/windows/memoryspaces) and RE-CODING. At every step.

With the right combination of cap+edit+store+transmit+display, you might not encounter much, if any, loss (e.g. preferred 3DBD format flow). Normally, however, there is SOME.

Surprisingly, of the 2 variants that you mentioned above (SbS & TaB), SbS has a worse track record in this than does TaB. This has to do with the fact that, ignoring capture for a moment and assuming the transmission doesn't change formats - quite possible with those 2 formats, there is no SbS-like display format (closest being checkerboard), while there is a TaB-like display format (being FPR-interleaved Circ. Polarized, aka "Passive"). Here's the details:

SbS has 960x1080 + 960x1080 (already 1/2 FullHD3D).
For Active displays, it gets unsqueezed into 1920x1080 + 1920x1080 (in each of 2 separate windows/memspaces), then those are alternated (at double speed) on output. You end up with slightly less than 1/2 quality of FullHD3D (because of the SbS storage and because of the interpolation due to unsqueezing). The framerate-per-eye is the same, even though the OVERALL framerate to the screen is doubled.

for Passive displays, it gets unsqueezed into 1920x1080 + 1920x1080 (just like in the last method), then alternating lines are removed from the L and the R (odd on the L, even on the R, or vice-versa depending upon manufacturer) giving 1920x540 + 1920x540, then the 2 views are displayed simultaneously (=1 interlaced view) at normal framerate giving 1920x1080. You end up with 1/2 quality for the source, a little less for the unsqueeze interpolation, and 1/2 for the line loss = 1/4 quality of FullHD3D.

conversely,
TaB has 1920x540 + 1920x540 (1/2 FullHD3D).
For Active displays, it gets unsqueezed into 1920x1080 + 1920x1080, then those are alternated (just like with SbS, except this time the unsqueeze is vertical instead of horizontal). You end up with slightly less than 1/2 quality of FullHD3D.

For Passive displays, it DOES NOT GET UNSQUEEZED. Instead, the lines that were consecutive are considered to be alternate/interlaced lines, and so it is combined interlaced onto the display with the other view. You end up with exactly 1/2 quality of FullHD3D (though there could be considered to be some artifacting when using pre-interpolated/squeezed lines as unsqueezed/interlaced alternate lines, so maybe you should say slightly less than 1/2 quality here too).


As you can see, SbS suffers much additional loss in one of those scenarios, whereas TaB suffers much additional loss in NONE of those scenarios.

However, SbS is often more commonly used in broadcast than TaB because SbS doesn't have problems with 1080i (interlaced) compressed transmission, while TaB does (relegating TaB use to mainly 720p transmissions). If you think about it, this makes sense.

Note: when using Passive screens, the alternating lines operate in a similar way to how interlaced 2D video works (although there is no temporal shift, and sometimes passive displays that use interlacing OF interlaced material can re-use the interlaced material by re-introducing it's complement in a 2nd field that becomes frame-duplicated when for 1080p).

Note Also: in NONE of the display formats is SbS nor TaB shown DIRECTLY. They are ALWAYS reformatted somehow.

Note Also, also: what I said way above in #3 of post #14 coincides with jagabo's answer.

Another Note: if you view S3D on common 3DTV displays WITHOUT 3D glasses, you will see either: Alternating frames of similar views on Active displays (giving a "wiggle 3D" shaking effect), or an interlaced "double image" of normal height/width and speed on Passive displays.

Does that answer all your questions?

Scott