Ratings section for 3D movies

[mAchiNE]

I think it would be great for there to be a rating system for 3D movies similar to what websites like IMDB do but the rating would not be for the content of the movie but a rating on the quality of the 3D in the movie, members on the forum who have seen the movie could rate the movies 3d out of 10 and the average rating be shown (I think this is how sites like IMDB operate for the ratings of the content/sory etc)

This way we can see what movies have the best 3D effects and not waste our time on the crap ones, also there could be a section for true 3D and converted 3D

[cybereality]

Thats an interesting idea actually.

[android78]

Great idea!
It would actually be nice to have several metrics that the community could vote on for the 3D movies though since different people like different things.
You could have:
- 3D depth (ranging from Virtually non-existent -> Real Life)
- Headache/vomit inducing factor (ranging from non -> Chunder Thunder!)
- Immersion (ranging from not at all -> Felt like I was there)
- Clever use (ranging from Meh -> Very Clever)
- Overall (ranging from Poor -> Perfect ... everyone should see it)

[Gae43]

It would be very worthwile.

I am currently watching a lot of 3D content on my new LG 3D Monitor and the 3D varies incredibly. Sometimes I'm watching 3D that looks phenomenal e.g. the LG 3D Demo clips and other content off YT3D, whereas other times I hardly notice the 3D effect e.g. Resident Evil Afterlife 3D. Here is a 3D movie that could have been stunning, but the 3D was sporadic throughout the film. At times it looks quite flat.
Films like Resident Evil Afterlife have a really negative effect on 3D fans like me. I have just spent lots of money and had lots of hassle buying a 3D monitor and then when I finally sit down with the glasses on, the reward is a film that has little 3D depth where the characters just have a slight edge to them in most scenes. That's not good enough. For most of the film, I had very little sense of the Z-depth and quite often was thinking, "What's the point of having a 3D monitor if this is what the 3D looks like?". I might as well watch it in 2D on my projector instead. That idea was quickly reversed though when I put on a clip of "Tangled" and saw just how beautiful it looked in 3D . Just the right combination of depth and bright vivid colours. The characters all looked rounded and there was a distinctive sense of space within each scene. Everything looked solid and you felt like you were looking onto real solid characters and objects (albeit animated ones). Hats off to the true artists at Disney for getting it right. It was so good, I forgot I was watching 3D but felt like I was looking on some new and unique form of animation, as though the animated objects were real and had just sprung to life. Does that make sense? It just looks so beautiful. I think out of all the 3D, even though "Tangled" may not have as much depth as some other material, overall it looks the most beautiful.

Gae43

P.S. Looking forward to watching my 3D Blu-ray rental of "Legend of the Guardians" at some point. From the few minutes that I briefly looked at earlier, both the animation and 3D looked excellent with great depth.

[Fredz]

That's a nice idea but the problem is people have a very personal and different appreciation of 3D, some people find 3D conversions acceptable when others can't stand them for example. So I'm not sure relying only on users rating would really be meaningful.

I think something more objective would be needed, in the same vein than GameGrade3D but for movies. To do that some objective metrics should be defined and measured in some side-by-side images extracted from the movies.

I think the most useful one would be the disparity range, which can be easily calculated for several meaningfull points in the images (foreground, zones of interest, background). I've done this several times with images from games and movies by simply calculating the distance between corresponding points in the left and right images using The Gimp.

Some examples :
- Measuring perceived depth in S3D games : http://www.mtbs3d.com/phpbb/viewtopic.php?f=3&t=13005" onclick="window.open(this.href);return false;
- Half-Life 2 : http://www.mtbs3d.com/phpbb/viewtopic.p ... 320#p67320" onclick="window.open(this.href);return false;
- Crysis 2 : http://www.mtbs3d.com/phpbb/viewtopic.p ... 342#p58342" onclick="window.open(this.href);return false;

Using this you can objectively compare disparity between 3D movies to see which ones will offer the more depth or which ones use pop-out. The depth range could also be calculated for the screen sizes that are used the most (24" monitors, 50" TVs, 100" projectors and 730" cinemas for example) with standard viewing conditions (around 30° FOV generally). You would then know how close and how far the scene would be in the virtual 3D space in each case.

Some other metrics based on these disparity calculations could also be used to detect errors, such as window violations, eye divergence, cardboard effects and vertical disparity.

I'm currently working on an application that automatically calculates a depth map and a 3D model from a side-by-side image, which could be useful in this regard. It's quite far from being ready for release - especially for live-action movies - but here is an example of what it's able to do on animated movies for now :

HTTYD-depth.png

[cybereality]

@Fredz: That program looks really interesting. I'd love to hear more about it as it progresses.

[Fredz]

Maybe I'll make a post about it, explaining how it does work and what are the problems I encounter so far. Still it's quite useful for comparing depth ranges for 3D movies, I'll post screenshots for other movies (Avatar, Priest, Tangled) when I'll be back to my dev machine. It shows pretty clearly how the 3D in Priest is a total joke for example.

[Likay]

Nice and interesting Fredz!
By a quick flyby thought the principle could be used to actually adjust the stereoscopic effect in a movie. Just take the depthmap and alter the brightness/contrast of it. As always with 2d+depth this will cause some issues with occluded zones (could actually be fixed by an enhanced algorithm since these are apparent in the original stereoscopic image) but it's nonethless fun playing with the thought. I know that the purpose of the program isn't about that though.

[Fredz]

By a quick flyby thought the principle could be used to actually adjust the stereoscopic effect in a movie. Just take the depthmap and alter the brightness/contrast of it. As always with 2d+depth this will cause some issues with occluded zones (could actually be fixed by an enhanced algorithm since these are apparent in the original stereoscopic image) but it's nonethless fun playing with the thought. I know that the purpose of the program isn't about that though.

Yep, this type of app could be used as a basis for depth modification through reprojection for 3D movies, like Crytek did for Crysis 2. As you said there will be problems with occluded zones, but state-of-the-art techniques in this research field (Middlebury Stereo Evaluation) are already able to reconstruct depth maps even for some occluded zones, so I guess the problem can somehow be minimized.

There's also an interesting paper explaining how to adapt depth for different viewing conditions based on the creation of a depthmap, but it seems to be quite complicated to get it right : Does size matters? The impact of screen size on stereoscopic 3DTV.

But the main problem is that these techniques are still far from being perfect, although still a lot better than what I do. The depth errors are quite low when compared to ground-truth, but they are still present and I think our eyes are very good at picking even small errors. Another problem is that sub-pixel precision isn't yet used as a metric in the Middlebury Stereo Evaluation, and that seems to be a lot more difficult to do correctly than only pixel precision according to my experiments.

Now to stay in the subject of this topic, here are some additionnal screenshots that show depthmaps and extracted 3D models for several 3D movies (perspective and side view). I think this illustrates quite well the depth range that can be perceived even if the calculations are quite bad most of the time.

How to Train Your Dragon (side view) :

HTTYD (side) - depth.png

Avatar (perspective view), good illustration of the problems I encounter with live-action :

Avatar - depth.png

Avatar (side view) :

Avatar (side) - depth.PNG

Priest (perspective view), good illustration of the cardboard effect (basically only 4 planes, no roundness for characters and only a flat plane for the background) :

Priest - depth.png

Priest (side view), good illustration of the very low depth compared to other movies :

Priest (side) - depth.png

[cybereality]

Ok, you win with Priest. Maybe the conversion was not as good as I thought. But it shows at least how easily people can be fooled with cheap conversions.

[Fredz]

Yep, you're quite right when you say that it's difficult for people to correctly appreciate the real depth of a 3D movie, I hope illustrations like this will help them understand why current 2D/3D conversions are nowhere near real 3D. This does also illustrate quite well why relying only on user comments is probably not a good idea for a 3D rating system for movies, some objective measures are obviously needed.

If people are interested I can try to post some more examples for other 3D movies.