HD Advisor September 3, 2010

[FXRH]

It is difficult to convert 2D to 3D successfully even when doing so on a frame by frame basis with a budget of millions. Expecting realistic results in a TV costing a couple of thousand dollars is ludicrous.

Let the name-calling begin! I figured it was about time for words like "ludicrous" to show up.

Many critics pointed out that the 2D-to-3D conversion of CLASH OF THE TITANS was poor. That conversion cost millions of dollars. I've seen bad CGI that was costly. I've seen Blu-ray movies on my "cheap" HDTV that look as good as or better than they did in a movie theatre with more expensive projection equipment. I've seen cheap TVs that look better than expensive TVs. So I don't find the "dollar amounts" argument entirely convincing.

It says nothing about the quality of the 3D the process produces however which varies from fairly good in some scenes to poor or nonexistent in others.

Thanks for granting that sometimes the process can yield "fairly good" results. I believe that's all I was claiming.

This discussion is starting to remind me of that Marx Brothers line: "Whom are you going to believe, me or your own eyes?" I saw a demo of the technology. It was better than I expected. There was plenty of room for improvement (as I said, the faux 3D gave itself away eventually) but I was surprised that it was decent in most of the scenes. To announce, without seeing, that the conversion process will largely yield poor results strikes me as being somewhat presumptuous.

I was impressed. My friend who owns a Samsung was impressed. Just for grins, I did a web search last night and found several reviews by others who were impressed. I don't know what the board policy is on providing links to other boards, so I won't post any, but I'll quote one paragraph from one review.

So does it work? Surprisingly, the answer is yes! We didn't hold out much hope for this function but we were surprised just how well it worked. The results are actually pretty good — impressive even — although they're still not a patch on the effect created by a genuine 3D source. We put it to the test with a selection of 2D films. To compare the 2D to the 3D version, all you need to do is close one eye, and when we gave it a go, most of us preferred the 3D version! It's actually quite surprising. Remember the chase scene in Casino Royale when Bond follows the villain to the top of a crane? Well, the crane really stood out from the sky, and chase scenes all flowed well and in the right direction as the cutting is cleverly done. This all works very well most of the time, but it's not perfect!

One thing that intrigued me from reading these reviews is that the results vary widely on different types of visual material and the technology fails almost completely on content such as drawn animation. Now that is interesting, and I can understand why the algorithms primarily designed to provide an "educated guess" on depth in a live-action film might come up with nothing on this kind of material.

Of course, there were also reviewers who were not impressed. But at least they viewed some conversion results before passing judgment.

And I note your comment that you have also seen the technology yourself. You obviously thought less of it than I did. But you did the fair thing: You saw first and judged later.

To those who haven't yet seen 2D-to-3D on-the-fly conversion: Please see it in action before announcing that it "largely" will not work. Who knows? You might be as surprised as I was.

[morphinapg]

Let the name-calling begin! I figured it was about time for words like "ludicrous" to show up.

Many critics pointed out that the 2D-to-3D conversion of CLASH OF THE TITANS was poor. That conversion cost millions of dollars. I've seen bad CGI that was costly. I've seen Blu-ray movies on my "cheap" HDTV that look as good as or better than they did in a movie theatre with more expensive projection equipment. I've seen cheap TVs that look better than expensive TVs. So I don't find the "dollar amounts" argument entirely convincing.

Not only was Clash of the Titans' 2d-to-3d conversion high budget, but it used MUCH slower processing (allowing for more detailed scanning), and it allowed people to tweak the 3d, or even completely modify it themselves if they wanted (If an object was in the wrong depth, they could set the depth themselves). Sure, this effect still suffers from the "cardboard cutout" effect I've been mentioning, but that's pretty much impossible to avoid without a real 3d image, but bigger budget conversion should produce MUCH more realistic depth, and that's not only because of the budget, but because of the techniques they use.

Sure, I've seen lower budget things outperform higher budget things as well, but even after the above explanation. A few thousand dollars vs millions of dollars? It's pretty obvious which would be of higher quality.

Thanks for granting that sometimes the process can yield "fairly good" results. I believe that's all I was claiming.

This discussion is starting to remind me of that Marx Brothers line: "Whom are you going to believe, me or your own eyes?" I saw a demo of the technology. It was better than I expected. There was plenty of room for improvement (as I said, the faux 3D gave itself away eventually) but I was surprised that it was decent in most of the scenes. To announce, without seeing, that the conversion process will largely yield poor results strikes me as being somewhat presumptuous.

First off, I've seen the technology. Not the exact technology these TVs are using, but similar technology. Second, if you know how the technology works and what its limitations are, you can easily know that it won't even come close to real 3d.

Also, guess what? I had these 3d glasses that would convert any color image into 3d. Yeah, they were cool, and they made things 3d, but the images were nothing like they should have been. Sure, some material worked better than others (for example, a Spider-man poster made spider-man pop out of the background, because red is the color that pops out the most) and some content looked horrible. Even when the content was generating all the wrong depth, it was still 3d, and so it might still fool some people, but it was completely wrong, and looked nothing like if the image was really 3d.

[cbcdesign]

Let the name-calling begin! I figured it was about time for words like "ludicrous" to show up.

Thanks for granting that sometimes the process can yield "fairly good" results. I believe that's all I was claiming.

I did not say you were ludicrous and therefore did not call you names. I said it was ludicrous to expect a two grand TV to convert 2D content to 3D with anything like the sort of fidelity a studio based system costing millions would do.

But that is the problem. "Sometimes" doesn't cut it, not for me anyway. If I went to a 3D movie at a theatre and the 3D varied from quite good in odd scenes to terrible in others and non existant in most, I would not be a happy consumer. I find it odd then that people seem content to accept this sort of performance in their expensive new 3D tv.

I tell you now, I am 100% certain that the vast majority of people will tire of the fadd that is 2D-3D conversion and will switch the feature off after a few months of use. It may not even take that long. I grew tired of it fairly quickly when I realised just how hit and miss it is.

[FXRH]

I did not say you were ludicrous and therefore did not call you names.

And I, in turn, didn't say you called me ludicrous, nor did I say that you called me names. I said that the word ludicrous appeared in your review. Characterizing an opinion as "ludicrous," even if no name is attached, strikes me as being needlessly provocative.

Who was "expecting" that on-the-fly real-time conversion would be as good as an expensive non-real-time one? This is a relatively new technology with much room for improvement; I didn't see anybody arguing otherwise.

When pejorative language starts showing up in a previously "frendly" disagreement, the discussion usually self-destructs fairly quickly. I just like keeping the tone civil. Silly me.

[gravis778]

But that is the problem. "Sometimes" doesn't cut it, not for me anyway. If I went to a 3D movie at a theatre and the 3D varied from quite good in odd scenes to terrible in others and non existant in most, I would not be a happy consumer. I find it odd then that people seem content to accept this sort of performance in their expensive new 3D tv.

I tell you now, I am 100% certain that the vast majority of people will tire of the fadd that is 2D-3D conversion and will switch the feature off after a few months of use. It may not even take that long. I grew tired of it fairly quickly when I realised just how hit and miss it is.

Just want to chime in. Its been a long weekend, otherwise I may have done it earlier.

First, let me start with the "Black and white" to "color" thing. Let me also claim that I am not an expert on this. First, in Black and White, many people miss-represent it - with the exception of very early days of photography, or very poor cameras, or bad reproductions, most works have different tones and shades of black or white. The shades and tones tend to be similar depending on the color that is being used. A lady's white face is not the same shade as white as her dress is in a picture. From that, you could probably derive a fairly good automated estimation of color. Its not going to be perfect, but its not like its going to substitute Yellow for Blue. That being said, you might not be able to get the exact color unless you built a profile for camera type, lense, film material and photographic chemicals used. Even then, it will probably be something that is "pretty darn good", but not great.

I could argue about 1080p upconverts. Its not nearly as good as true HD. So why do we turn the feature on? Because it usually DOES look better than the 480i or p source material. Worst case scenario is that you see little to no difference between it and the source material. I have never seen it look worse.

I don't think anyone here will argue that the 2D to 3D technology is going to look anywhere near as good as true 3D material or something converted. I think the Cardboard cutouts example, though, is flawed. I think that is worse case scenario. With just the slightest bit of motion in an object, you could probably render a tad bit of depth to the image. It may look okay, even neat, and in some cases even halfway decent. I truthfully really want to see this in action. I just haven't seen a store do it yet - they are always playing the demo discs.

[Josh Z]

First, let me start with the "Black and white" to "color" thing. Let me also claim that I am not an expert on this. First, in Black and White, many people miss-represent it - with the exception of very early days of photography, or very poor cameras, or bad reproductions, most works have different tones and shades of black or white. The shades and tones tend to be similar depending on the color that is being used. A lady's white face is not the same shade as white as her dress is in a picture. From that, you could probably derive a fairly good automated estimation of color. Its not going to be perfect, but its not like its going to substitute Yellow for Blue. That being said, you might not be able to get the exact color unless you built a profile for camera type, lense, film material and photographic chemicals used. Even then, it will probably be something that is "pretty darn good", but not great.

Two problems with this:

1) Because there are so many more possible colors than there are shades of gray, a lot of different colors often wind up translating to the same shade of gray. Any sort of automated software trying to reverse-engineer the colors would have no way of determining which of the multiple color options any particular shade of gray is supposed to map to. In a best case scenario, you'd wind up with a very cartoonish color picture that's all primaries.

2) When filming black and white movies, the costume designers and art directors work with the camera people to choose colors that translate to the best black & white tonalities. Often, very outlandish colors are needed to get appealing grays. So, even if a software really could accurately reverse-engineer the colors, you'd find out that Cary Grant's immaculate wool suit is actually bright green and he's standing in front of pink wallpaper!

[morphinapg]

I don't think anyone here will argue that the 2D to 3D technology is going to look anywhere near as good as true 3D material or something converted. I think the Cardboard cutouts example, though, is flawed. I think that is worse case scenario. With just the slightest bit of motion in an object, you could probably render a tad bit of depth to the image. It may look okay, even neat, and in some cases even halfway decent. I truthfully really want to see this in action. I just haven't seen a store do it yet - they are always playing the demo discs.

Again you are misinformed. The cardboard cutouts example is truthfully the only way this technology is even currently possible, and then it's still not even close to the real depth whatsoever. Even the big budget movies can't get away from the cardboard cutout issue. Theoretically, it might be possible to extract further depth from individual objects, but that would require so much processing power that it wouldn't be worth it anyway.

Assuming this technology works based on motion (the only way I could possibly see it "working") the only scenes it would "work" on are scenes with good parallax, aka the camera moving left or right, or up or down. Any other scene, like a zooming shot or a static shot wouldn't even come close to working. Also, a lot of "parallax" shots often move quite slow, just slow enough for you to notice. Those shots probably wouldn't work too good either.

However, you do lose picture quality with these methods, because they distort the original picture information to achieve the 3d effect, as well as losing brightness in your tv when wearing the 3d glasses. The comparison to upscaling is a bad one. Upscaling leaves 100% of the original picture information intact (unless it's a bad upscaler with too much post processing, etc) whereas 2d-to-3d loses information due to the distortion of the original image, as well as losing the original cinematic intent by the director.

[gravis778]

Again you are misinformed. The cardboard cutouts example is truthfully the only way this technology is even currently possible, and then it's still not even close to the real depth whatsoever. Even the big budget movies can't get away from the cardboard cutout issue. Theoretically, it might be possible to extract further depth from individual objects, but that would require so much processing power that it wouldn't be worth it anyway.

Assuming this technology works based on motion (the only way I could possibly see it "working") the only scenes it would "work" on are scenes with good parallax, aka the camera moving left or right, or up or down. Any other scene, like a zooming shot or a static shot wouldn't even come close to working. Also, a lot of "parallax" shots often move quite slow, just slow enough for you to notice. Those shots probably wouldn't work too good either.

However, you do lose picture quality with these methods, because they distort the original picture information to achieve the 3d effect, as well as losing brightness in your tv when wearing the 3d glasses. The comparison to upscaling is a bad one. Upscaling leaves 100% of the original picture information intact (unless it's a bad upscaler with too much post processing, etc) whereas 2d-to-3d loses information due to the distortion of the original image, as well as losing the original cinematic intent by the director.

Sorry, I may have rambled a bit instead of getting to the point. My thought of 2D to 3D conversions I was thinking of is something like NASCAR. Constant camera movements, constant moving of the cars. You could probably render a passable 3D car, especially if its coming around a corner coming toward the camera, and not have it look like a cardboard cutout. However, its speculation. I don't know how the technology works, and would LOVE if someone could post a link.

In any case, I get that its a gimick, and that some stuff will look like cardboard cutouts. I saw Last Airbender. Some scenes were quite good, others were quite flat. But other movies are quite well done - such as Alice In Wonderland. Granted, those were movies that techs tweaked, but they were also probably putting stuff into a 3D wireframe field on a 20 inch monitor, then hitting render.

In any case, my question was answered - has anyone seen it, and how good was it. The answer seems to be, pretty good, but not perfect. That is actually better than I was expecting, and strengthened my reasons for getting a 3DTV.

[morphinapg]

Sorry, I may have rambled a bit instead of getting to the point. My thought of 2D to 3D conversions I was thinking of is something like NASCAR. Constant camera movements, constant moving of the cars. You could probably render a passable 3D car, especially if its coming around a corner coming toward the camera, and not have it look like a cardboard cutout. However, its speculation. I don't know how the technology works, and would LOVE if someone could post a link.

In any case, I get that its a gimick, and that some stuff will look like cardboard cutouts. I saw Last Airbender. Some scenes were quite good, others were quite flat. But other movies are quite well done - such as Alice In Wonderland. Granted, those were movies that techs tweaked, but they were also probably putting stuff into a 3D wireframe field on a 20 inch monitor, then hitting render.

In any case, my question was answered - has anyone seen it, and how good was it. The answer seems to be, pretty good, but not perfect. That is actually better than I was expecting, and strengthened my reasons for getting a 3DTV.

As for Alice in Wonderland (which I have not seen, in 3d or 2d) I believe that movie was made with 3d conversion in mind, which means they would have adjusted their camera movements, tweaked the 3d themselves, any compositing would be done in 3d space, and any CG effects would probably be real 3d.

As for your Nascar reference, I think Nascar (and most sports) would probably not be very good for 2d-to-3d conversion, as the camera does not usually physically move. The camera stays in place and pans back and forth, as well as zooming in and out. Because the camera stays from one perspective, there is no true parallax. Unless each car was moving at exactly the same speed and the camera was very close to the action (not just zoomed in, but physically close), it's unlikely the 3d effect would look good, if it even worked at all for the footage.

By the way, with B&W to Color conversion, there approximately 65,000 colors per shade of grey.

[Josh Z]

As for Alice in Wonderland (which I have not seen, in 3d or 2d) I believe that movie was made with 3d conversion in mind, which means they would have adjusted their camera movements, tweaked the 3d themselves, any compositing would be done in 3d space, and any CG effects would probably be real 3d.

Yes, both Alice in Wonderland and Piranha 3D were shot with 2-D cameras (in the case of Piranha, because shooting on and under the water caused a host of issues with 3-D cameras), but were designed for 3-D conversion from the start. Both movies had 3-D specialists that worked in pre-production and on the set to ensure that the photography was optimized for 3-D conversion.

As opposed to either Clash of the Titans or The Last Airbender, which were quickie 3-D conversions done at the last minute.

Even so, a lot of viewers complained about the cardboard cut-out look in both Alice and Piranha.

[gravis778]

As for Alice in Wonderland (which I have not seen, in 3d or 2d) I believe that movie was made with 3d conversion in mind, which means they would have adjusted their camera movements, tweaked the 3d themselves, any compositing would be done in 3d space, and any CG effects would probably be real 3d.

As for your Nascar reference, I think Nascar (and most sports) would probably not be very good for 2d-to-3d conversion, as the camera does not usually physically move. The camera stays in place and pans back and forth, as well as zooming in and out. Because the camera stays from one perspective, there is no true parallax. Unless each car was moving at exactly the same speed and the camera was very close to the action (not just zoomed in, but physically close), it's unlikely the 3d effect would look good, if it even worked at all for the footage.

By the way, with B&W to Color conversion, there approximately 65,000 colors per shade of grey.

Okay, so it just occured to me this morning that, if TV manufactorors are including this technology in TVs, there are probably software solutions out there that I could try this with. I found http://www.tridef.com/home.html which offers a 14 day free trial. I loaded it onto the laptop with anaglyph turned on (ugh, yes, flimsy red / blue glasses, I already have a headache) just to see how it works.

I loaded up some old home videos that I had shot out at the lake a few years back. The result - better than I expected. While its deffinately not true 3D, it was fun, and certainly better results than the "cardboard cutouts" scenario. That being said, nothing popped out of the screen, but rather the screen was like a window with depth behind it. I was honestly shocked that the water looked so good and that it was able to do real-time 3D conversion. The stuff shot on the shore was even more impressive.

This software seems to work with different systems, so I might just head out and pick up some nVidia active shutter glasses later, and use my MPC to display movies on my TV until I get around to picking up a real 3D tv later.

Oh, and I did some research. Yes, the 3D effect on most 2D to 3D is done by motion. I noticed on the videos I shot with a handheld camera, the ones that were slightly more shaky had better 3D than the ones that were shot with a steady hand or on a tripod. Great, most of my videos are shot with a hand-held camera!

The software also does photos, and the 3D on those is REALLY weird. It did seem to work MUCH better with digital photos than with scanned photos though.

Yes, I am VERY pleased with the results that this software produced. Its sad that $50 software and anagryph glasses seemed to produce better 3D in a real-time conversion than what was done in Last Airbender.

[AV_Integrated]

Josh - As an FYI at Infocomm both Sony and Hyundai were showing polarized LCD displays.

The Sony was their own brand and was outrageously expensive for a 40" model. It was color calibratable for studio monitor work and was priced around $10,000 I believe.

The Hyundai, interestingly, was a Sharp LCD with a black piece of tape over the Sharp logo. Not sure if the guts were their own or Sharp's, but I'm guessing that it was actually a Sharp display.

That sucker was a 65" model and looked REALLY good. I would take a set of passive glasses over active shutter glasses if the price was right in a second. If Sharp were to release a 65" display and it allowed for use of passive glasses I would be far more interested in that instead of their active glasses models. Hyundai would NOT discuss MSRP with me at all. I let them know how thrilled I was with that.

I can see a point where full 1080p LCD displays can support passive glasses and maintain resolution. It would be amazing to see that sooner rather than later at a reasonable price. I'm just not sure there wil ever be a price piont on that which is attainable vs. the active solutions.

[Lee Stewart]

Josh - As an FYI at Infocomm both Sony and Hyundai were showing polarized LCD displays.

The Sony was their own brand and was outrageously expensive for a 40" model. It was color calibratable for studio monitor work and was priced around $10,000 I believe.

The Hyundai, interestingly, was a Sharp LCD with a black piece of tape over the Sharp logo. Not sure if the guts were their own or Sharp's, but I'm guessing that it was actually a Sharp display.

That sucker was a 65" model and looked REALLY good. I would take a set of passive glasses over active shutter glasses if the price was right in a second. If Sharp were to release a 65" display and it allowed for use of passive glasses I would be far more interested in that instead of their active glasses models. Hyundai would NOT discuss MSRP with me at all. I let them know how thrilled I was with that.

I can see a point where full 1080p LCD displays can support passive glasses and maintain resolution. It would be amazing to see that sooner rather than later at a reasonable price. I'm just not sure there wil ever be a price piont on that which is attainable vs. the active solutions.

Hyundai has been selling a 46" 3D LCD (passive) in Japan for a couple of years now. I believe the MSRP is $4600. It like all the other passive LCDs use Xpol technology.

Next year, Vizio will release their 65" passive 3DTV. All they said about the price was . . . "it will be affordable."

[gravis778]

I can see a point where full 1080p LCD displays can support passive glasses and maintain resolution. It would be amazing to see that sooner rather than later at a reasonable price. I'm just not sure there wil ever be a price piont on that which is attainable vs. the active solutions.

If its a polorized display, though, wouldn't you have to double the resolution to achieve 1080p? I agree, I would take polorized over active shutter any day, but I don't think I would do it at the cost of cutting my resolution in half. So wouldn't you require almost a 4k display to render a 1080p polorized picture? Its no wonder that its costly

[morphinapg]

If its a polorized display, though, wouldn't you have to double the resolution to achieve 1080p? I agree, I would take polorized over active shutter any day, but I don't think I would do it at the cost of cutting my resolution in half. So wouldn't you require almost a 4k display to render a 1080p polorized picture? Its no wonder that its costly

You would only need double the resolution in one dimension, aka 3840x1080 or 1920x2160

EDIT: But they might do 4k anyway to allow 2d 4k signals.