Likely a different master or different preprocessing was applied.Instead of all the guessing going on, can you just ask Warner if that scene looked identical on the master used as to the VC-1 encode and let us know? Have you asked them?

--Darin

Elephant's Dreams is still being downloaded to my computer, but I was able to find a few stills from the movie at this site:

http://orange.blender.org/media-gallery

In examining the stills, I believe this one will give VC-1 the most problems:

http://orange.blender.org/wp-content/themes/orange/images/media/gallery/s2_bridge.jpg

There is computer shading here that contains some banding, and the question is how much an encoder will degrade this. The image is around 254 kB (jpeg).

If there is anyone on this forum whose specialty is image processing, I'd appreciate some input on the following theory. In every image there is a certain amount of entropy. Why? Because not every pixel is captured or rendered with maximum (perfect) focus. Sometimes this is intentional. Thus the information you see in the defocused and blurred areas is less than the focused areas, and contain a form of correlated noise from adjoining areas in the frame. VC-1 may attempt a form of hyper-resolution in extracting or presenting more information than is present in the original picture. Or, its interpretation of same. We all know that the algorithm is highly efficient because it produces incredibly sharp encodes with few bits (heck, it might even reach the Shannon limit). A related theory I have is that the heuristics within VC-1 may be overaggressive, when it should be softening the image to preserve the entropy in the master.

Elephant's Dreams is still being downloaded to my computer, but I was able to find a few stills from the movie at this site:

http://orange.blender.org/media-gallery

In examining the stills, I believe this one will give VC-1 the most problems:

http://orange.blender.org/wp-content/themes/orange/images/media/gallery/s2_bridge.jpg

There is computer shading here that contains some banding, and the question is how much an encoder will degrade this. The image is around 254 kB (jpeg).
As you'll see when your download is complete, the VC-1 version shows substantially less banding than the JPEG you linked to (I transcoded my source .avi from the 8-bit PNG sequence).

As you'll see when your download is complete, the VC-1 version shows substantially less banding than the JPEG you linked to (I transcoded my source .avi from the 8-bit PNG sequence).

I agree. This movie is computer-generated, so we are dealing with a lot of textures and shading that make it difficult to judge the aesthetics of patches and banding after encoding. There is even simulated smoke which makes it harder to judge. The VC-1 encode looks pretty good though.

If there is anyone on this forum whose specialty is image processing, I'd appreciate some input on the following theory. In every image there is a certain amount of entropy. Why? Because not every pixel is captured or rendered with maximum (perfect) focus. Sometimes this is intentional. Thus the information you see in the defocused and blurred areas is less than the focused areas, and contain a form of correlated noise from adjoining areas in the frame. VC-1 may attempt a form of hyper-resolution in extracting or presenting more information than is present in the original picture. Or, its interpretation of same. We all know that the algorithm is highly efficient because it produces incredibly sharp encodes with few bits (heck, it might even reach the Shannon limit). A related theory I have is that the heuristics within VC-1 may be overaggressive, when it should be softening the image to preserve the entropy in the master.
I think you mix up entropy and redundancy. Entropy is what remains if you remove the redundancy. The basic information that can not be further compressed without losing some information. All video compression used on HD-DVD and BR is lossy and not near the entropy of the footage (except for trivial noise free cases) (is that what you mean by Shannon limit?). Lots of information is removed. The art of video compression is to remove only the information whose removal the HVS won't notice.

I agree. This movie is computer-generated, so we are dealing with a lot of textures and shading that make it difficult to judge the aesthetics of patches and banding after encoding. There is even simulated smoke which makes it harder to judge. The VC-1 encode looks pretty good though.
Ah, but you can compare it to the source in this case!

http://media.xiph.org/ED/ED-1080-png.torrent

That's the uncompressed source frame sequence I generated the .avi files from.

I'm going to post this comparison again because... it's sunny outside. ;) Thanks, KS. I was hoping for the links to where this was originally posted. Are these captures that you made, or were they posted elsewhere originally? Do you know how these captures were made?

They're derived from the shots that were included in post #1 of this thread. Check it out.

How can a film master have banding?

Because it was filmed with a digital camera.

Which I could understand if the digital camera in question was 6bit(or lower) in colour resolution.

But it's said because the colour resolution(which was in 12bit) had to be resampled to the higher colour resolution of HD formats.

Doesn't make sense to me to be honest, but I'm no expert.

How you can remove gradient detail by processing, some kind of soften filter?

I think they only way this can be resolved if someone gets in contact with someone that has worked on or seen the master.

How you can remove gradient detail by processing, some kind of soften filter?Yeah, so I think "H.264 can smooth out banding in a master! VC-1 shows the banding as is!" sounds none other than BS. I'd like to know which super intelligent human-like codec can process it.

Yeah, so I think "H.264 can smooth out banding in a master! VC-1 shows the banding as is!"
Look at my sample yet?

sounds none other than BS. I'd like to know which super intelligent human-like codec can process it.
It's at least as likely earlier in the processing chain as in the codec itself. And there is already at least one compression tools on the market with an integrated debanding filter.

... And there is already at least one compression tools on the market with an integrated debanding filter.

Ben, which one? I only became interested in the idea recently and didn't know they were out there. I'd even considered writing one for ffdshow decoding.

- Tom

Ben, which one? I only became interested in the idea recently and didn't know they were out there. I'd even considered writing one for ffdshow decoding.

I probably shouldn't say, since I'm not sure if they're marketing that feature, although I know it's deplyed in at least one post house.

...I did this at home on my own gear, so I'm putting it up as a .torrent.
http://216.99.212.230:6969/torrents/Elephants_Dream_WMV.torrent?9C267F763BFE9570D5C439ADD8B6D0D7 3C7708E6Thanks, Ben.
It's an impressive result considering 96% of the "info" is discarded (the resulting file size is 4% of the original)!

I have a few questions regarding this encode:

1. Why 24fps? Is this how the original was designed?
I thought you mentioned that 24/60 fps in real life always means 23.976/59.94.

2. Why no Lookahead?
I think zambelli om doom9 claimed that this improves quality, although slows down the encode.

3. Are the used encoding parameters most suitable for animated CGI or will they be in the "optimal settings" category for other footage as well (e.g. OTA hidef recordings)?

Thanks.
Diogen.

Thanks, Ben.
It's an impressive result considering 96% of the "info" is discarded (the resulting file size is 4% of the original)!
And the source was already lossless (PNG) encoded. It's more like 1-2% compared to the uncompressed source.

I could drive the bitrate down substantially without hurting quality much if I spent some time with it. I really only sweated the gradient demo part with this encode.

I have a few questions regarding this enode:

1. Why 24fps? Is this how the original was designed?
I thought you mentioned that 24/60 fps in real life always means 23.976/59.94.[/QUOTE]
For whatever reason the original project was rendered at 24.00 fps, and I left it at that so I didn't have to resample the audio.

2. Why no Lookahead?
I think zambelli om doom9 claimed that this improves quality, although slows down the encode.
Lookahead helps a lot for 1-pass encodes, but doesn't do anything useful with 2-pass encodes (you can think of Lookahead as offering a little bit of 2-pass in a 1-pass encode).

3. Are the used encoding parameters most suitable for animated CGI or will they be in the "optimal settings" category for other footage as well (e.g. OTA hidef recordings)?
I'm pretty happy with these as general defaults. The two I'd be most likely to tweak are DQuant and Perceptual Option - those can have quite source dependent presets (and are two of the main things that get varied per-segment with PEP/PSE).

Thanks, Ben.
Really appreciated.

Diogen.

Look at my sample yet?


It's at least as likely earlier in the processing chain as in the codec itself. And there is already at least one compression tools on the market with an integrated debanding filter.And why is that debanding filter applied only to H.264 and not VC-1? You are just deliberately replacing the question with a slightly different one.

And why is that debanding filter applied only to H.264 and not VC-1? You are just deliberately replacing the question with a slightly different one.
There are plenty of ways to deband source before VC-1 encoding.

I'm sure this conversation has gone on for many times longer than the compressionist spent on that one shot :).

Is there an actually a torrent linked above? I tried clicking on it, and I get no torrent file downloaded. Strange. I've tried two different browsers now.

Is there an actually a torrent linked above? I tried clicking on it, and I get no torrent file downloaded. Strange. I've tried two different browsers now.
I see three people downloading it right now.

Is this the URL you're using?

http://216.99.212.230:6969/torrents/Elephants_Dream_WMV.torrent?460640F9C1B039A20FF60AFED5F2F02D A4B00598

Yikes, I see what's happening - a new file got added accidentally and messed up the hash.

Try this link now. I'll lock the directory.

http://216.99.212.230:6969/torrents/Elephants_Dream_WMV.torrent?9C267F763BFE9570D5C439ADD8B6D0D7 3C7708E6

For whatever reason the original project was rendered at 24.00 fps, and I left it at that so I didn't have to resample the audio.
I can't quite make sense of it myself. The audio tracks (available on the project website) are longer than the video, even when the video is set to 23.976 fps. They don't provide much information about the intended playback speed.

Lookahead helps a lot for 1-pass encodes, but doesn't do anything useful with 2-pass encodes (you can think of Lookahead as offering a little bit of 2-pass in a 1-pass encode).
Well, I wouldn't say "a lot". It helps with detecting scene changes, flash frames and fades in 1-pass. 2-pass takes care of all that automatically.

Well, I wouldn't say "a lot". It helps with detecting scene changes, flash frames and fades in 1-pass. 2-pass takes care of all that automatically.
Hey, I didn't know you were posting here!

(we work together on the codec team).

As he says, a better qualification would have been "helps a lot with some kinds of content, particularly flashes, fades, and rapid editing." Trying to do real-time encoding of movie content, this is a pretty big win. Doesn't do much for a webcam, though.

ben,

Thanks a ton for the really cool encode. For some reason while watching, I couldn't help wonder what it would take to make that video a 3x HD DVD? I know you most likely don't have a ton of time, but it would be really cool if we could turn this encode into a sort of open source group authoring experiment. Adding simple HDi menus and the like, so we can all see/learn how an HD DVD is made.

It would be fun to burn this to a DVD and hav 3x functionality (simple menus). How much more work would that be?

PS-Also, I am a total noob. What are the other files in the torrent? A reg file? At first I thought maybe this WAS to make it a 3x burn, but I guess not....

ben,

Thanks a ton for the really cool encode. For some reason while watching, I couldn't help wonder what it would take to make that video a 3x HD DVD? I know you most likely don't have a ton of time, but it would be really cool if we could turn this encode into a sort of open source group authoring experiment. Adding simple HDi menus and the like, so we can all see/learn how an HD DVD is made.

It would be fun to burn this to a DVD and hav 3x functionality (simple menus). How much more work would that be?
Probably not possible - I made that finally mainly to demo the gradiant thing, not to make it strictly HD DVD compatible.

PS-Also, I am a total noob. What are the other files in the torrent? A reg file? At first I thought maybe this WAS to make it a 3x burn, but I guess not....
Registry keys. You can look at it in any text editor.

Probably not possible - I made that finally mainly to demo the gradiant thing, not to make it strictly HD DVD compatible.

That's fine. I am just really curious about making 3x discs with menus, where can I go to find that stuff out then?

Registry keys. You can look at it in any text editor.

So the registry file changes ones computer to do certain specific things during a vc-1 encode, I assume.

What are the other two files for?
ED_avi_convert.prproj
Elephant's Dream 10-25.wme

Hey, I didn't know you were posting here!
Join date Feb 2004. :) It's just been difficult to get a word in! :D Nah, just kidding, I was on a little hiatus from AVSF but I think it's time to get back into it.

That's fine. I am just really curious about making 3x discs with menus, where can I go to find that stuff out then?
Ulead has some software for this. Apple's DVD Studio Pro can as well (albeit MPEG-2 only).

So the registry file changes ones computer to do certain specific things during a vc-1 encode, I assume.
Exactly. The .png shows the settings from Zambelli's tool.

What are the other two files for?
ED_avi_convert.prproj
Elephant's Dream 10-25.wme
.prproj = Premiere Pro project, which I used to convert the .png sequence to a .avi source file
.wme = The Windows Media Encoder session file.

For those who like comparisons, the following links are Zambelli's (VC-1 at 6Mbps) & Saggitaire's (AVC & MPEG2 at 6,12,18 Mbps) ElephantDream encodings. They are at 25fps(?) and restrickted by the theoretical minimum (video at 29.4Mbps max) HD-DVD limitations, (although i think the 12&18Mbps encodings violate the HD-DVD vbv limits).
"?ttp://jfl1974.free.fr/upload/Elephant/""?ttp://multimediacom.free.fr/HD-DVD/H264/""?ttp://multimediacom.free.fr/HD-DVD/MPEG2/"
*put h instead of ?

For those who like comparisons, the following links are Zambelli's (VC-1 at 6Mbps) & Saggitaire's (AVC & MPEG2 at 6,12,18 Mbps) ElephantDream encodings. They are at 25fps(?) and restrickted by the theoretical minimum (video at 29.4Mbps max) HD-DVD limitations, (although i think the 12&18Mbps encodings violate the HD-DVD vbv limits).
"?ttp://jfl1974.free.fr/upload/Elephant/""?ttp://multimediacom.free.fr/HD-DVD/H264/""?ttp://multimediacom.free.fr/HD-DVD/MPEG2/"
*put h instead of ?
Note that these are not entirely apples-to-apples. There are some differences in buffer constraints and motion vector search range between the VC-1 and AVC.

Likely a different master or different preprocessing was applied.

"Let's work the problem people. Let's not make things worse by guessing."

;)

Engineers rock!

btw - I just started watching Planet Earth. I was totally unaware of the issue and it nearly made me jump off my chair!

Found this on the BD software forum, I don't know if it's been posted over here:

http://i10.photobucket.com/albums/a127/wgrose/IMG_0804.jpg

Well Ben you pushed me to read again the doom9 thread.

Ben wrote,
"Note that these are not entirely apples-to-apples. There are some differences in buffer constraints and motion vector search range between the VC-1 and AVC."

As about the 6Mbps encodings it is really apples to apples between VC-1/AVC.
The buffer was set (faultly) to 9800KB for both,
The 25fps setting helped the free Ms encoder not been in disadvantage (relative to x.264)because of its non adaptive B frames number, so both used 15 frames GOP. If it was 24fps the non "adaptive B frames" VC-1 would use GOP 12 and the adaptive x.264 GOP 14.
The mv range was set at 1024 horizontal for both and vert. 512(avc) or 256 (vc-1) which is already overkill. Setting mv range to "auto" makes the encoder to use rather smaller ranges.

What i would like for a visual comparison is the restrickted 6Mbps vc-1 encode vs your unrestrickted 10Mbps. This will give us an indication of the point of diminissing returns on this type of content. Pitty that i am blocked out of torrents.

Oh .. by the way. Are there any updates for the free WMV9 encoder on shedule. Adaptive B frames would be a nice feature.

Found this on the BD software forum, I don't know if it's been posted over here
Do you think it is the camera causing the banding, or maybe the odd screen orientation?

Do you think it is the camera causing the banding, or maybe the odd screen orientation?

Neither.

If you have this disc, the banding in that scene would be extremely obvious to you.

For what it's worth, the MPeg versions of Planet Earth had loads of banding in the scenes referenced in this thread - would seem to suggest confirmation that it's in the masters.

...in your opinion.

For what it's worth, the MPeg versions of Planet Earth had loads of banding in the scenes referenced in this thread - would seem to suggest confirmation that it's in the masters.

or the MPEG2 compression is also causing banding.

Well, if two codecs show clearly visible banding and a third codec shows noticably less banding, this allows two different interpretations:

(1) Either the excessive banding is caused by the two codecs, while the third codec is doing better at this scene.
(2) Or the banding is in the master and the third codec somehow hides the banding. In this case the two codecs showing the banding are nearer to the source.

It's virtually impossible to know which interpretation is correct without seeing the master. However, if the banding "locations" are exactly identical between the two codecs which show the excessive banding, that would suggest (it would still not be a proof, though) that the banding is more likely to be in the source. If the location of the banding is different in the two codecs, it's more likely that the banding was caused by the codecs.

Well Ben you pushed me to read again the doom9 thread.

Ben wrote,
"Note that these are not entirely apples-to-apples. There are some differences in buffer constraints and motion vector search range between the VC-1 and AVC."

As about the 6Mbps encodings it is really apples to apples between VC-1/AVC.
The buffer was set (faultly) to 9800KB for both,
The 25fps setting helped the free Ms encoder not been in disadvantage (relative to x.264)because of its non adaptive B frames number, so both used 15 frames GOP. If it was 24fps the non "adaptive B frames" VC-1 would use GOP 12 and the adaptive x.264 GOP 14.
The mv range was set at 1024 horizontal for both and vert. 512(avc) or 256 (vc-1) which is already overkill. Setting mv range to "auto" makes the encoder to use rather smaller ranges.
Sorry, I was thinking of the earlier encodes.

That said, I don't believe Alex used DQuant or Adaptive Deadzone for those encodes, which help a lot with this clip. I did a 5 Mbps version of Elephant's Dream that look quite good as well.

What i would like for a visual comparison is the restrickted 6Mbps vc-1 encode vs your unrestrickted 10Mbps. This will give us an indication of the point of diminissing returns on this type of content. Pitty that i am blocked out of torrents.
Those wouldn't be apples to apples either.

Instead of worrying about existing encode, can you tell me what it is you'd like to see here - might be easiest to just make a new sample.

Oh .. by the way. Are there any updates for the free WMV9 encoder on shedule. Adaptive B frames would be a nice feature.
Nothing since the fall v11 release. We did announce our commercial VC-1 SDKs, but thos are implemented as a library, not a .dll, and so won't support Windows Media Encoder.

Ben thanks for the info.

Although no news -> bad news.

I don't care to make an apples to apples comparison. I 'll try to see the difference between apples and oranges. If you can cut a clip of your 10Mbps encode (frames 2500 to 3500) and upload somewhere (except torrents) together with the settings you used it will be fine.

Sorry if I am a little off topic.

I just watched The Fountain, and it has two very obvious instances of posterization.

00'35"
http://img517.imageshack.us/img517/5682/fountain01ly5.jpg

08'33"
http://img228.imageshack.us/img228/2136/fountain02hdjf3.jpg

I saw this movie at the theater and it didn't have that.

Note: the picture noise has been caused by the high ISO used and wasn't in the image.

How many time do they need to tell us over and over? It's in the MASTER! Only VC1 can give us all in the MASTER! But there is still hope (if the new format follows DVD path), someday there maybe a RE-master re-release. (IMO, not on this title and I don't case the movie!)

regards,

Li On

Wow, Grubert! Definitely no way to argue banding isn't there, in those pics! Ironically, not even the dither of your high ISO setting was able to save it. ;)

I just watched The Fountain, and it has two very obvious instances of posterization.

Note: the picture noise has been caused by the high ISO used and wasn't in the image.
That high ISO makes it really hard to determine what's really going on in that image in any diagnostic sense. The comparison to theater also isn't always useful, since the release print will tend to smooth out banding.

Anyway, I think I've adequately demonstrated at this point that VC-1 is perfectly capable of handling even challenging gradients.

That high ISO makes it really hard to determine what's really going on in that image in any diagnostic sense. The comparison to theater also isn't always useful, since the release print will tend to smooth out banding.

Anyway, I think I've adequately demonstrated at this point that VC-1 is perfectly capable of handling even challenging gradients.

Yes, but the various samples here suggest that sometimes it does not. It would be nice to know why so we can predict under which conditions it becomes a problem.

- Tom

I just received my copy of the fountain from blockbuster online and wanted to check out those two instances of specific scenes. There is indeed some banding going on there. In both cases, the bitrate dropped down very low. For instance, the first scene at 0'35" dropped to about 3.8 mbps while the other scene at 8'33" was around 3 mbps - 6 mbps. I don't remember such obvious banding in the theatrical version. Warner did something wrong with this encode.

Yes, but the various samples here suggest that sometimes it does not. It would be nice to know why so we can predict under which conditions it becomes a problem.
What is so hard to understand about that? There can be a number of reasons why that can happen. And they have all been mentioned multiple times already. Let me give you a list again:

(1) The banding could be in the digital master.
(2) A bad 10bit to 8bit conversion could be in fault.
(3) The 2 pass encoder could have allocated too few bits on this scene.

Even if it's (3), you still can't say that it's a general fault of VC-1. This can and does happen with every encoder. It's the duty of the compressionist to check the 2 pass encoding and to correct any such problems in the 3rd pass (or 4th or 5th).

Anyway, it would be nice, if there was some kind of "quality police", which would try to find the cause of such image quality issues, so that they can be avoided in the future. But such a thing wouldn't work without very active participating of the studios and compressionists. And I don't think there's much chance that this will happen.

What is so hard to understand about that? There can be a number of reasons why that can happen. And they have all been mentioned multiple times already. Let me give you a list again:

(1) The banding could be in the digital master.
(2) A bad 10bit to 8bit conversion could be in fault.
(3) The 2 pass encoder could have allocated too few bits on this scene.

Even if it's (3), you still can't say that it's a general fault of VC-1. This can and does happen with every encoder. It's the duty of the compressionist to check the 2 pass encoding and to correct any such problems in the 3rd pass (or 4th or 5th).


Maybe it's a fault of VC1 and maybe not. But I'll wager MSFT will find a way to improve it if we continue to point out shortcomings of movies that just, coincidently, happen to have been encoded in VC1.


Anyway, it would be nice, if there was some kind of "quality police", which would try to find the cause of such image quality issues, so that they can be avoided in the future. But such a thing wouldn't work without very active participating of the studios and compressionists. And I don't think there's much chance that this will happen.

That's why they need help from enthusiastic and knowledgeable consumers. It's called complaining and it gets a lot of things fixed. And doing it online together at places like AVS is more fun and has more leverage. If a product is vulnerable in some fashion they have to live with it being publicly pointed out.

Usually, telling those vocal online consumers to just "shut up" does not work well.

- Tom

That high ISO makes it really hard to determine what's really going on in that image in any diagnostic sense.

Okay, 50 ISO now.

http://img115.imageshack.us/img115/9170/fountain1v2mv4.jpg

http://img515.imageshack.us/img515/3331/fountain2v2jr8.jpg

http://img407.imageshack.us/img407/2453/fountain3v2oc8.jpg

Player is a Toshiba HD-A1. Display is a Sony VW50 1080p projector.

The comparison to theater also isn't always useful, since the release print will tend to smooth out banding.



OK. So we have the possibility that Efilm (http://www.efilm.com/filmography.php) screwed up the digital intermediate by adding banding, but the banding then disappeared when the prints were struck.

Or we have the alternate hypothesis (reinforced by donricouga) that the bitrate was simply insufficient, and that, as with Planet Earth, insufficient bitrate manifested itself in the form of posterization.

Occam's razor.

...but what about MS's razor? :p

I think ben said endless time, bitrate has NOTHING to do with any of VC1 problem. VC1 has magic (I mean tweak!).

On the other hand, MPEG2 on blu-ray has some banding scene as I already pointed out. Any definite banding scene on a blu-ray AVC title? Some said Open Season and I don't see any issue besides a perfect picture on my setup.

regards,

Li On

OK. So we have the possibility that Efilm (http://www.efilm.com/filmography.php) screwed up the digital intermediate by adding banding, but the banding then disappeared when the prints were struck.

Or we have the alternate hypothesis (reinforced by donricouga) that the bitrate was simply insufficient, and that, as with Planet Earth, insufficient bitrate manifested itself in the form of posterization.

Occam's razor.
Even if it were an issue with insufficient bitrate at a particular moment, that's not an issue with VC-1 intrinsically, but an operator error since they could easy up the bitrate for that shot.

That said, gradients normally encode quite nicely. DQuant + Adaptive Dead Zone tends to handle them very nicely without much bitrate impact.

Note we now have Region of Interest for encoding, so gradient-friendly settings can be applied to just the gradient portion of the image, leaving the rest of the image to use the default mode.

Okay, 50 ISO now.

Yep, looks like banding. I didn't see that in the theater nor have I heard anything specific about compression for that particular title. Neither frame looks at all challenging, though, for our VC-1 encoder. Either it was a source issue, or some unusual non-standard settings were applied.

Thanks for the inputs Ben.

Im doing gradients test (test different gradients etc to provoke the encoder)

And Im never getting this severe banding even at AVG bitrates 1mbit-2mbit (1080/23,976P) wih either AVC or WMV. So I have really hard time belive the encoder would have problem at 5mbit-6mbit.

But what i would like to ask is. Why is it common in some masters? Wouldnt they have better tools when they do the 12bit/10bit - 8bit conversion?

So the encoder is the culprit? Is there another non-MS encoder for VC-1?

Yep, looks like banding. I didn't see that in the theater nor have I heard anything specific about compression for that particular title. Neither frame looks at all challenging, though, for our VC-1 encoder. Either it was a source issue, or some unusual non-standard settings were applied.

Ben -

I'm not an expert on VC1 parms but it seems possible there is some default or awfully tempting option on automatic 2-pass that allows this sort of banding to occur when they don't have enough quality control. (like watch the darn thing)

Possibly you could just post some warnings and/or better defaults for the lazier members of the encoding community.

- Tom

Just a thought, but maybe some of these problems are with the transfer method between the player and the display. The HDMI input on my Panny 50PHD7UY shows posterization on discs that are simply not there through the component input, so much so that I use the component out of the HD-A1 exclusively. If it is a problem with the encode or on the madter, wouldn't I see it on both sets of inputs. Both inputs and all scan rates have been ISF'd so there is a level playing field.

Just a thought, but maybe some of these problems are with the transfer method between the player and the display. The HDMI input on my Panny 50PHD7UY shows posterization on discs that are simply not there through the component input, so much so that I use the component out of the HD-A1 exclusively. If it is a problem with the encode or on the madter, wouldn't I see it on both sets of inputs. Both inputs and all scan rates have been ISF'd so there is a level playing field.


I'd like to hear more reports on this, if HDMI causes posterization on some setups.

- Tom

That's why they need help from enthusiastic and knowledgeable consumers. It's called complaining and it gets a lot of things fixed. And doing it online together at places like AVS is more fun and has more leverage. If a product is vulnerable in some fashion they have to live with it being publicly pointed out.

Usually, telling those vocal online consumers to just "shut up" does not work well.
The danger with that is that too much complaining may drive insiders away. I know that I wouldn't have much fun posting here, if all that people did was complaining about my product. Even if it was justified. But if it was not justified, it'd be even more frustrating cause I couldn't do much to improve the situation.

Anyway, I'm not at all against complaining if there's something wrong. I just hope that we don't drive people away from the forum by pushing too hard. That has happened before. Frustrating insiders does not help anybody. Ok, 'nough said!

Thanks for the inputs Ben.

Im doing gradients test (test different gradients etc to provoke the encoder)

And Im never getting this severe banding even at AVG bitrates 1mbit-2mbit (1080/23,976P) wih either AVC or WMV. So I have really hard time belive the encoder would have problem at 5mbit-6mbit.


For reference the first shot is a fade-in from black onto a rather stationary shot of a sheet where the biblical quote is being written.

The problem may be that studios have believed that it was really safe to go below 10 Mbit/sec average as advertised months ago (http://www.avsforum.com/avs-vb/showthread.php?t=718689) (along with a 20-30% decrease in peak bitrate required). This, coupled with the overall good performance of VC1 with a lot of titles, may have made them overconfident.

At any rate, there was a shocking lack of QC. After all, a simple user could spot it.

The problem may be that studios have believed that it was really safe to go below 10 Mbit/sec average as advertised months ago (http://www.avsforum.com/avs-vb/showthread.php?t=718689) (along with a 20-30% decrease in peak bitrate required). This, coupled with the overall good performance of VC1 with a lot of titles, may have made them overconfident.
If the compressionist just dialed 10 Mbit/s average in and never looked again at the final image quality, he simply didn't do his job properly and should be fired. But I don't really believe that to be the case.

Of course once we arrive at a situation where there are hundreds of discs released every months by one studio, the compressionists might not have the time to check every frame for problems. They might then just use the automatic 2 pass encoding and limit their manual work on doing some control sampling. But I don't think we're at this stage yet.

At any rate, there was a shocking lack of QC. After all, a simple user could spot it.
And that is exactly the best argument for searching the bug in the source. The compressionist must have been very much sleeping to miss such a strong banding. So the best explanation would be that the compressionist had no chance to do better because the banding was already in the source. As you may imagine, it's probably not in the authority of a compressionist to order a remastering. A compressionist will probably have to do the best with the what he is given. And if there's banding in the source, he can't help it.

Of course I'm only guessing. But a totally sleeping compressionist sounds less likely to me than banding in the source.

The problem may be that studios have believed that it was really safe to go below 10 Mbit/sec average as advertised months ago (http://www.avsforum.com/avs-vb/showthread.php?t=718689) (along with a 20-30% decrease in peak bitrate required).Unfortunately, a fair number of unsuspecting people here were misled there. Forunately, compressionists are unlikely to be as easily misled by things like that.

--Darin

I'd like to hear more reports on this, if HDMI causes posterization on some setups.

- Tom
I've seen this with regular DVD players. LGA-531 upconvert DVD player has nasty posterization through HDMI with some DVDs, but not when using component output. The same scene will also look normal with some software players, and posterize with others. Therefore I think posterization may also be effected by how the players and software handle gradients and color space changes. Perhaps this is something that can be corrected through FW upgrade.

All i know is Grubert saw posterization on his setup, HD-A1 and a 1080p projector and i've seen the same thing on my PS3 and my Sammy. So i guess there are two possibilities. The encoding was starved due to insuffient data, or trbarry thinks it could be HDMI. Well we can test that by simply having component cables plugged in instead of HDMI. I don't have any so if someone does, then please chime in.

I've seen this with regular DVD players. LGA-531 upconvert DVD player has nasty posterization through HDMI with some DVDs, but not when using component output. The same scene will also look normal with some software players, and posterize with others. Therefore I think posterization may also be effected by how the players and software handle gradients and color space changes. Perhaps this is something that can be corrected through FW upgrade.

Since you can't upconvert with component cables, I don't think this comparison works well against what we are trying to find out.

I reported the exact problem LONG ago. If it's not in the source, then maybe Warner intentionally screwing VC1 and back stab Microsoft? Don't ask me why though! :D

Btw, just finished "Apocalypto". There is a looking-into-the-Sun fade out scene (during the eclipse scene). The AVC bitrate is around 30Mbps at the time and NO banding whatsoever! :cool:

regards,

Li On

I reported the exact problem LONG ago. If it's not in the source, then maybe Warner intentionally screwing VC1 and back stab Microsoft? Don't ask me why though! :D

Btw, just finished "Apocalypto". There is a looking-into-the-Sun fade out scene (during the eclipse scene). The AVC bitrate is around 30Mbps at the time and NO banding whatsoever! :cool:

regards,

Li On

Very nice ! Maybe Disney wanted to outdo Warner in regards to their crappy sun shot in Planet Earth :p
BTW, I haven't seen this movie yet. Is it worth a blind buy ?

All i know is Grubert saw posterization on his setup, HD-A1 and a 1080p projector and i've seen the same thing on my PS3 and my Sammy. So i guess there are two possibilities. The encoding was starved due to insuffient data, or trbarry thinks it could be HDMI. Well we can test that by simply having component cables plugged in instead of HDMI. I don't have any so if someone does, then please chime in.


I'm not claiming HDMI is the problem here. It just caught my attention above and I'd heard other comments about it so I thought we should try to eliminate that as one possible source if we could get enough data points.

Besides, I have a pet peeve against HDMI simply because of its incompatibilities and copy protection, but that's probably for another thread. ;)

- Tom

Well, the picture and sound are great but the movie may not suit everyone. I suggest a rental first if you have the option.

regards,

Li On

I just watched The Fountain, and it has two very obvious instances of posterization.

00'35"
http://img517.imageshack.us/img517/5682/fountain01ly5.jpg

08'33"
http://img228.imageshack.us/img228/2136/fountain02hdjf3.jpg

I saw this movie at the theater and it didn't have that.

Note: the picture noise has been caused by the high ISO used and wasn't in the image.
I specifically looked for this last night during my viewing of The Fountain. On my setup, it did not have any banding. I have HD-XA1 connected to 1365 x 1024 D-ILA projector via component output.

Good for you as now you found a perfect reason to upgrade your hometheater setup.

regards,

Li On

This brings up an interesting point.

Does anybody know what color space(s) their displays use? And what their HDMI connection is sending.

The native on-disc color space is component (YCbCr), correct? And displays need to eventually obtain RGB to drive their panels.

So where is this conversion happening? In the deck? Is HDMI transmitting RGB or YCbCr? If the latter, then the display must then convert to RGB, correct?

YCbCr-to-RGB conversion can cause color error and loss of dynamic range if done incorrectly I believe. (As I think was evidenced by some problems caused by displays that had DVI inputs where adapters were used). Do we know for a fact that these conversion processes are not part of the issue?

How big are the file sizes of these movie files with bandings? Do they fill 30GB in HD DVD?

Also, for those who see bandings over HDMI, do they disappear when you connect it via an analog connection? You can test it without much hassle.

A wrongly calibrated display can also cause banding.

If that were the case, they would be seeing bandings in numerous movie titles they happen to play on their system. Occam's Razor asks why there are these isolated cases of banding appearing.

Until we see actual direct screenshot captures there's too many variables involved.

If that were the case, they would be seeing bandings in numerous movie titles they happen to play on their system. Occam's Razor asks why there are these isolated cases of banding appearing.

Perhaps we don't know they are not?

I also don't know if a specific conversion error would only manifest itself with specific content.

Perhaps we don't know they are not?


Three years with a BenQ PE8700 (single-chip 720p DLP projector) made me acutely sensitive to posterization. ;)

First time posterization annoyed me (not counting old mpeg1 stuff) was on my DVD version of Crimsom tide. But I always assumed, this is gonna be the downside of going from analog to digital. But its nice the process evolves and its gonna be a less of a problem.

I think the best way would be that the companys maked sure the 8bit masters are free from banding. Filter if they have to.

If that were the case, they would be seeing bandings in numerous movie titles they happen to play on their system. Occam's Razor asks why there are these isolated cases of banding appearing.For HDMI too.

According to Hidef Digest, Letters from Iwo Jima has banding as well. As expected, it is a VC1 title from Warner.

Hopefully we will see the screenshots of the 4 instances of banding that was seen in this release.

link
http://hddvd.highdefdigest.com/lettersfromiwojima.html


The source is pristine and there isn't any noise, crushing, or artifacting even when bright, swirling smoke veils a shot. The only problems with the visual presentation are minor. First, I caught four instances of faint color banding -- the most noticeable of which occurs early in the film when Watanabe gazes out across at the gray sky (I should point out was staring at the light skies specifically to see if the problem would pop up -- most people won't even catch the banding at all, and it certainly didn't detract from the extremely impressive video presentation). Second, like the video presentation on 'Flags of Our Fathers,' some of the seams of the special effects are exposed by the high-def presentation. Again, this isn't the fault of the transfer per se, and with time I'm sure that the effects houses will catch up with the heightened scrutiny that comes with high definition.

Getting back to the original topic of this far-to-long thread, I finally got a chance to speak to an industry insider who has seen the source and encodes for both Happy Feet and Planet Earth. They confirmed that the banding was "at least 90% source."

With that, I think I'll wrap up my own participation in this thread - we've got insider information on two of the movies in question, and a downloadable clip with repro steps and available sources. I'm not sure what I could contribute beyond that.

Any more info on that h.264 feature that can mask banding?

Any more info on that h.264 feature that can mask banding?
It's actually not a codec feature, it's a preprocessing feature.

I'm not sure what I could contribute beyond that.

You could tell them to fix their sources for next time. :)

Any ideas what the 10% were? Encoding issue?

I guess that is the best we can expect. Thanks for trying Ben!

BTW, there may be a bit of banding in The Matrix. However, it will be a while before I can verify my filter chain is working properly. If you are curious, you can look at any scene in the Matrix where there is bright light (for example, Neo being lifted out of the water into the ship).

You could tell them to fix their sources. :)

(I guess that is the best we can expect. Thanks for trying!)

And that's where we come in. We must act as very squeaky wheels and blow the whistle every time it is required. It may very well be that the compressionists actually spot the problem areas but are overruled by the mastering department. ;)

If nobody hollers at the theater when the projector is out of focus, the projectionist may take very long to focus it. ;)

I agree Grubert. Eternal vigilance! :D

You could tell them to fix their sources for next time. :)
Were it only that easy. My Photoshop-using friends know of what I speak, and you've got to do it 24 times a second!

Any ideas what the 10% were? Encoding issue?
"At least 90%" :)! That's not the same as saying that 10% was the codec.

BTW, there may be a bit of banding in The Matrix. However, it will be a while before I can verify my filter chain is working properly. If you are curious, you can look at any scene in the Matrix where there is bright light (for example, Neo being lifted out of the water into the ship).
That kind of CG/compositing heavy movie that predates the HD formats is exactly where you'd expect to see a lot of banding in the source.

for "Happy Feet" posterization/quality comparison: here are some high-quality still shots, resized to 1920x817px with Photoshop CS2 (bicubic, saved as .png) from 4096x1743px .jpg-images (sized ~1-2mb) ... if somebody can use them ;) ... if not, they're quite some eye candy and nice to watch :D

http://www.imagebanana.com/img/dj08le1/happyfeet_uhq_png_01.png

http://www.imagebanana.com/img/xuggf3d9/happyfeet_uhq_png_02.png

http://www.imagebanana.com/img/k9nabx14/happyfeet_uhq_png_03.png

http://www.imagebanana.com/img/anzftcsx/happyfeet_uhq_png_04.png

©2006 Warner Bros. Entertainment Inc. - U.S., Canada, Bahamas & Bermuda.
©2006 Village Roadshow Films (BVI) Limited - All Other


if you want more, tell me ... some are still left ;)

Those shots are better than the BRD/HD-DVD shots. We got screwed. ;)

Those shots are better than the BRD/HD-DVD shots. We got screwed. ;)

Yes were are those 468mbit discs

Actually, you could probably get that kind of quality with avc or vc-1 on today's formats. The reason why I said it looks better than what we got is that these pics have more detail and pop than the actual movie, at least presented on BRD or HD-DVD.

Where did these shots come from?

[...]
Where did these shots come from?
these are some of the promotional pictures offered by Premiere (German Pay-TV) promoting the pay-per-view-broadcast of "Happy Feet" ... ;)

There would seem to be plenty of potential left for improvement given the program sizes that did end up on br/hdvd for this movie (12.2 and 13.1 GB, respectively). Of course, there is the alternate theory by benwaggoner, that once you have "all the bits you need", adding more won't achieve any gains. That would seem to imply that vc-1 simply cannot achieve any better pq as we currently observe on those discs. You be the judge... ;)

Maybe we should have a vote to see what people think should happen if the master has visible banding which isn't supposed to be there. I know there is a school of thought that the compression tool's job is only to output what is input, but if the compressionist could reencode sections so that they look better than the master (fix the artifact) as far as banding, I would vote for that and consider that capability to be an advantage of a particular encoder.

--Darin

Maybe we should have a vote to see what people think should happen if the master has visible banding which isn't supposed to be there. I know there is a school of thought that the compression tool's job is only to output what is input, but if the compressionist could reencode sections so that they look better than the master (fix the artifact) as far as banding, I would vote for that and consider that capability to be an advantage of a particular encoder.

--Darin

Im all for a correction of Banding before encoding. I dont want an automated function in the encoder that does it. T2 has learned us to never trust the machines. ;)

Of course, there is the alternate theory by benwaggoner, that once you have "all the bits you need", adding more won't achieve any gains. That would seem to imply that vc-1 simply cannot achieve any better pq as we currently observe on those discs. You be the judge...

Of course improvements can be made. The underwater shot (with the banding) could obviously use more attention (bits). I'm sure there are other scenes that would benefit.

Of course improvements can be made. The underwater shot (with the banding) could obviously use more attention (bits). I'm sure there are other scenes that would benefit.

If you add more bits to encode a banding that is in the master, the only thing that will happen is that the encoder will make the banding more true to the banding in the master.

If you add more bits to encode a banding that is in the master, the only thing that will happen is that the encoder will make the banding more true to the banding in the master.

You may be right but that would mean if you add more bits to encode a part that has no banding, it will be more true to the master as well ;)

Can someone compare the frames that msv posted to the same frames from the HD-DVD/Blu-ray disc? I wonder if the blue sky has banding on the disc versions. If so, then we know the real source shouldn't have it. :)

P.S. msv I'd love to have more shots! Then we can compare against the discs and pick them apart. ;) (ESPECIALLY underwater shots)

That brings us back to the necessity of being specific and precise in what we say. For instance, when benwaggoner says his connection tells him the banding was 90% the fault of the source, is that referring to:

...the movie as a whole? (too non-specific)

...the specific scenes of interest at this forum? (can't get anymore direct than that)

...other scenes in the movie having entirely no relation to the instances of banding we are discussing here? (a complete misrepresentation of the context of interest here)

The wording of benwaggoner's explanation could fall into any one of those scenarios and still be true. However, the answer we are seeking can be remarkably different amongst those 3 situations (essentially, which one is "more true").

Hanky

I finally got a chance to speak to an industry insider who has seen the source and encodes for both Happy Feet and Planet Earth. They confirmed that the banding was "at least 90% source."

Thats seems like he refering to the scens we discuss.

No, it is equally vague, once you get right down to it. We could assume the industry insider is referencing the particular scene, but that simply cannot be discerned directly from the statement.

Im all for a correction of Banding before encoding. I dont want an automated function in the encoder that does it. T2 has learned us to never trust the machines. ;)I don't want an automated function (at least not one that happens automatically based on thinking that there is banding that shouldn't be there). Fixing problems before encoding is fine with me, but baring that, I want one that the compressionist can use on a specific segment when they see problems. If it didn't make the section look better, they could always go back.

--Darin

You may be right but that would mean if you add more bits to encode a part that has no banding, it will be more true to the master as well ;)

It all depends on what type of scene. But after a certain point, just adding more bits will make so small improvment nobady can se them.

Just an example (not scientific)

If the source is 100% PQ

a 5mbit encode will be 50% PQ
a 10mbit encode will be 60% PQ
a 15mbit encode will be 65% PQ
a 20mbit encode will be 67% PQ
a 25mbit encode will be 68% PQ
a 30 mbit encode will be 68,5% PQ
a 35mbit encode will be 68,75% PQ

So it all depends on how much difference a human eye can detect(24fps HD).

If its on 1% basis then it cant detect the difference between 25mbit and 35mbit.
If its on 2% basis it cant se difference between 20mbit and 35mbit

So dont take the numbers for real, but that basicly how it works. Films are lossy encoded for human eye vision, not for computer analys.

To take it up a level of complexity, the fall-off for improvement may be entirely variable, based on the nature of that scene (how much hf detail, noise load, amount of movement/change). The knee of the curve could be 15...could be 25...could be 45...could be 100 Mb/s. If it's 15, then naturally going to 20 is not going to net much, if any, improvement. If it is 35, then certainly there will be commensurate improvement going from 15-20, 20-25, 25-30, all the way up to 35.

[...]
P.S. msv I'd love to have more shots! Then we can compare against the discs and pick them apart. ;) (ESPECIALLY underwater shots)
no problem ... here are the next 4 (4 more left) ... unfortunately there aren't any more underwater shots available :( ;)

http://www.imagebanana.com/img/qhtbbeon/happyfeet_uhq_png_05.png

http://www.imagebanana.com/img/89cxt7m/happyfeet_uhq_png_06.png

http://www.imagebanana.com/img/59zi39ot/happyfeet_uhq_png_07.png

http://www.imagebanana.com/img/m3j35kg2/happyfeet_uhq_png_08.png

©2006 Warner Bros. Entertainment Inc. - U.S., Canada, Bahamas & Bermuda.
©2006 Village Roadshow Films (BVI) Limited - All Other

In my opinion, banding is a very bad and distracting artifact. I fully understand that it may be in the source, that it may not appear on the super-duper-better-than-reference CRT monitors that video compressionists are using for quality control. Nevertheless, most people will be watching the end result on digital displays that are not calibrated and/or don't have perfect video processing. Those involved in the encoding process should take the extra step and view the end result on an average LCD display, check the gradient scenes, and do whatever it takes to minimize or eliminate banding. Excuses of the sort "It's in the source" or "It doesn't appear on our reference display" are too convenient, and I'm sure the professionals realize that.

My. Goodness. Those shots make the hd-dvd/blu-ray happy feet look like it was a VHS recording. If anyone wants to tell me lowbitrate VC-1 is transparent, i would call them a downright liar. Shame we are told such lies.

My. Goodness. Those shots make the hd-dvd/blu-ray happy feet look like it was a VHS recording. If anyone wants to tell me lowbitrate VC-1 is transparent, i would call them a downright liar. Shame we are told such lies.

The images wasnt from the master.

and here are the final 4 images ;)

http://www.imagebanana.com/img/2dgaum98/happyfeet_uhq_png_09.png

http://www.imagebanana.com/img/yssexv4u/happyfeet_uhq_png_10.png

http://www.imagebanana.com/img/k00mj9b/happyfeet_uhq_png_11.png

http://www.imagebanana.com/img/bt5lrhh/happyfeet_uhq_png_12.png

©2006 Warner Bros. Entertainment Inc. - U.S., Canada, Bahamas & Bermuda.
©2006 Village Roadshow Films (BVI) Limited - All Other

msv ownz us all! :D

The images wasnt from the master.

Do you work for warner? If not, then you have no business saying that. Those are scenes from the movie, the movie i paid quite a bit over the DVD price to get in HD, and now i know for a fact that we (the consumer) were given a bitstarved and overall crappy HD transfer by warner bros while a VERY high quality master exists and not taken advantage of because of the use of VC-1 and such low bitrate. I am very upset.

Can anyone get the e-mail address of Warner bros so i can demand a return of my money on this disc until they put out something worth my money?

I could swear there were more HF pix, but here is one from hdvd, from the BR/HD/broadcast screenshots topic: (it's an underwater shot, so not quite comparable to these 2k samples- no pun intended ;) )

http://www.imageviper.com/displayimage/81877/0/Happyfeethddvd1.png
http://www.imageviper.com/displayimage/81879/0/Happyfeethddvd3.png

msv-

what is the source for those images?

Transparent, or no? You be the judge. :D

msv-

what is the source for those images?
the presslounge of our German Pay-TV broadcaster Premiere ... these pictures were published accompanying the "Happy Feet" pay-per-view-broadcast ... and I assume the pictures originally do come directly e.g. from Warner Bros. Germany (the rights holder) along with the broadcast-rights-package ;)

the presslounge of our German Pay-TV broadcaster Premiere ... these pictures were published accompanying the "Happy Feet" pay-per-view-broadcast ... and I assume the pictures originally do come directly e.g. from Warner Bros. Germany (the rights holder) along with the broadcast-rights-package ;)

Very interesting. Thanks for posting them!

Xylon! We need you!!

MSV, can you post one full-sized pics so we can see what's "beyond HD"?

MSV, can you post one full-sized pics so we can see what's "beyond HD"?
sure ... which of the 12 do you want? ;)

happyfeet_uhq_png_09.png looks like it could be interesting.

The HD version looks really soft next to these pics.

happyfeet_uhq_png_09.png looks like it could be interesting.

The HD version looks really soft next to these pics.
okay here it is ;) ... if you want more, just say it :)

http://www.mbmg.de/images/happyfeet_original_jpg_09.jpg

©2006 Warner Bros. Entertainment Inc. - U.S., Canada, Bahamas & Bermuda.
©2006 Village Roadshow Films (BVI) Limited - All Other

Do you work for warner? If not, then you have no business saying that. Those are scenes from the movie, the movie i paid quite a bit over the DVD price to get in HD, and now i know for a fact that we (the consumer) were given a bitstarved and overall crappy HD transfer by warner bros while a VERY high quality master exists and not taken advantage of because of the use of VC-1 and such low bitrate. I am very upset.
Those are digital production stills, NOT frames from the movie. Note that there is no motion blur. If you played that kind of frame at 24p, they wouldn't look at all film-like.

There's a reason why press pictures are taken by a press photographer, instead of just blowing up a frame from the 35mm print :).

Wow that's truly stunning!

I wish Warner would encode with VC-1 the way BVHE and Interscope have done.

If they are posed still renders why are they 2.40:1? Wouldn't they be in a little more print friendly AR? Just wondering.

For someone who has the movie can you confirm that these are actual frames from the movie? Maybe they are but without motion blur.

Those are digital production stills, NOT frames from the movie. Note that there is no motion blur. If you played that kind of frame at 24p, they wouldn't look at all film-like.

There's a reason why press pictures are taken by a press photographer, instead of just blowing up a frame from the 35mm print :).
ah okay ... I also got this non-OAR picture (http://www.mbmg.de/images/happyfeet_original_jpg_13.jpg) which seems to have motion blur, hasn't it? :)

preview, click here for full size (http://www.mbmg.de/images/happyfeet_original_jpg_13.jpg)
http://www.mbmg.de/images/happyfeet_preview_jpg_13.jpg

Should there be motion blur in these scenes where each penguin is essentially standing in place and moving their beak to coincide with the dialogue?

It's not unreasonable that these pics are derived from actual frames from the movie, right?

There was an IMAX release that probably took advantage of the 4k version, if a 4k version was actually rendered.

I dunno...benwaggoner seems to be implying that a layer of this extreme pq from generated stills must inherently be sacrificed in whatever processes are involved to generate a digital moving picture master.

It's also possible, that since these stills come from very close to the source in high-quality png form, that we may be seeing effectively 4:2:2 level performance which is showing to be a cut above the standard 4:2:0 level imagery of the hdm formats?

Just wondering -- how much do you think HD video disc buyers are willing to pay to make sure that every frame (24 per second) is HD "perfect"?
good point ,it does get alittle nit-picky around here doesnt it

I dunno...benwaggoner seems to be implying that a layer of this extreme pq from generated stills must inherently be sacrificed in whatever processes are involved to generate a digital moving picture master.
It's not quality being sacrificed, it's that an optimal still doesn't have motion blur, and an optimum movie frame does. Each is the right thing for its context.

In my opinion, banding is a very bad and distracting artifact. I fully understand that it may be in the source, that it may not appear on the super-duper-better-than-reference CRT monitors that video compressionists are using for quality control. Nevertheless, most people will be watching the end result on digital displays that are not calibrated and/or don't have perfect video processing.

I don't think so. IF the god awful ugly in your face banding are really in the source, then those "super-duper-better-than-reference CRT monitors" should only make the banding more obvious, otherwise those "profressional" monitor is not that true to the master then.

regards,

Li On

So someone posted photographic still images to compare to the movie screen captures? Sneaky! But hopefully it was not intended as a deliberate deception.

Wow. If not for Ben on this forum, we would have another source of misinformation no doubt! I wonder who could be behind this? ;)

When you shoot with an "ordinary" camera you will have shutterpeed of 1/48 that creates the filmmotion you se in most films.

If you set shutterspeed on 1/500 every frame gonna have very little motionblur, that will create the effect you have seen in Saving Private Ryan and Gladiator(some scenes) were the motion seems very unatural.

If you dont add motion blur to a animated movie they will have the saving privat ryan look. And thats not only the actors moving, the smooth camera pans will not be so smoth.

So for you that have a digital still camera and can adjust the shutter. Take a picture the same time you panning yourself(and the camera) and test with taking shutter 1/50 and 1/500 you will se the difference very clear.

What is the source of this screengrabs? HD discs or something else?

What is the source of this screengrabs? HD discs or something else?
something else ;)

those pictures were provided for promotional purposes by Premiere (German Pay-TV broadcaster) accompanying the PPV-broadcast of "Happy Feet" and I assume they come along as part of the "broadcast rights package" given to the TV stations by the rights holder ... but what the original source is I unfortunately don't know :)

something else ;)

those pictures were provided for promotional purposes by Premiere (German Pay-TV broadcaster) accompanying the PPV-broadcast of "Happy Feet" and I assume they come along as part of the "broadcast rights package" given to the TV stations by the rights holder ... but what the original source is I unfortunately don't know :)

Thanks. Thats all I need to know.

It's not quality being sacrificed, it's that an optimal still doesn't have motion blur, and an optimum movie frame does. Each is the right thing for its context.

Like I said earlier, if it is simply the case of motion blur, then there should be negligible effect to the static shots presented, since everybody in the image is just standing still and talking. There's no argument, that detail has to be sacrificed when motion blur is applied to a scene that contains motion. What we are observing is something other than that.

So either your theory has to concede that adding motion blur causes an inherent pq loss in all scenes (whether there is motion or not), or it has to accomodate that there is an inherent pq loss in the compilation process that creates a "moving picture encoding" from these stills. That is the only real "out" to explain how the vc -1 version seems to be a noticeable cut below these still samples, while maintaining that it is still "transparent" to the moving picture master.

We haven't seen the actual VC-1 stills of these scenes, have we? I'd like to compare them.

And that's where we come in. We must act as very squeaky wheels and blow the whistle every time it is required. It may very well be that the compressionists actually spot the problem areas but are overruled by the mastering department. ;)

If nobody hollers at the theater when the projector is out of focus, the projectionist may take very long to focus it. ;)

Absolutely right.

Wow - I can hardly wait for those 4K holographic discs! :p

Yes, these 4K stills do look very sharp, but as has been stated above, they are rendered still shots, and so do not have motion blur.

Motion blur is rendered into the actual motion picture before they ever get to the encoding phase, and it is NOTY to save bandwidth, but rather to make the motion appear more natural. You wouldn't want Happy Feet to look like Saving Private Ryan, which was shot on high speed film deliberately to enhance the graphic, gritty feeling of out of control battle.

Motion blur has also been implemented on the PS3 to try to make games look more realistic, I might add.

Also, it goes without saying that just because banding is not present in these stills, doesn't mean it is not present in the HD masters. These stills did not come from the HD master.

I realize that some folks will not understand this concept and will continue a one-track pursuit.

Wait, hold on a sec everyone... Motion blur? Shutter speed? Photo stills? Ummm, wasn't "Happy Feet" computer animated? None of those concepts would actually exist in the production process except when artificially simulated.

Wait, hold on a sec everyone... Motion blur? Shutter speed? Photo stills? Ummm, wasn't "Happy Feet" computer animated? None of those concepts would actually exist in the production process except when artificially simulated.
Of course motion blur is simulated in CG animation, otherwise you'd experience very jerky movement.

Watch "Monster House", it uses little to no motion blur and ends up looking like stop motion animation.

Wait, hold on a sec everyone... Motion blur? Shutter speed? Photo stills? Ummm, wasn't "Happy Feet" computer animated? None of those concepts would actually exist in the production process except when artificially simulated.

The talking penguins wouldn't exist in the production process except when artifically simulated too. I'm not sure I understand your point...
Shutter speed is simply a parameter of the CGI "camera" and the software will automatically render any appropriate motion blur on objects depending on the amount of motion and the shutter setting.

rdjam, consider that the banding occurs on objects that are not motion-blurred. Like the sky or the water in the background. Also consider that blurring should eliminate the banding.

I'm not sure why you didn't consider the above. Good thing there are other points of view that are more plausible. One-track mind indeed. :)

I'm not sure I understand your point...
My point was that it's a digital source and it's somewhat of an exception to the typical film-to-digital workflow. Some people were wondering about how it's possible to get this or that kind of quality from the movie or how ultraquality screenshots could exist. In the case of a CGI movie, there are nearly infinite ways to render the video.

Given that the new HDMI 1.3 seems to make a big deal about Deep Color:
http://www.hdmi.org/pdf/HDMI_Insert_FINAL_8-30-06.pdf
Deep Color
HDMI 1.3 supports 30-bit, 36-bit and 48-bit (RGB or YCbCr) color depths, up from
the 24-bit depths in previous versions of the HDMI specification.
• Lets HDTVs and other displays go from millions of colors to billions of colors
• Eliminates on-screen color banding, for smooth tonal transitions and subtle
gradations between colors
• Enables increased contrast ratio
• Can represent many times more shades of gray between black and white. At 30-bit
pixel depth, four times more shades of gray would be the minimum, and the
typical improvement would be eight times or more
and the fact the the Toshiba HD-XA2 player went up to using a 12 bit video DAC for component video output -- I wonder how much of the banding being reported is display related?

Using component video connection from a XA2 to a CRT based display would seem to give one 36 bit color display capability right now.
Are not "Deep Color" HDMI 1.3 LCD displays due to be available soon or even now?
http://www.tacp.toshiba.com/televisions/product.asp?model=52LX177

I just watched The Fountain, and it has two very obvious instances of posterization.

00'35"
http://img517.imageshack.us/img517/5682/fountain01ly5.jpg


I saw this movie at the theater and it didn't have that.

Note: the picture noise has been caused by the high ISO used and wasn't in the image.

That is the exact same on the region B version which is a BD50 Mpeg2 the bitrate at that scene is 34mbps. It would seem this is not a VC1 issue on this movie.

Given that the new HDMI 1.3 seems to make a big deal about Deep Color:
http://www.hdmi.org/pdf/HDMI_Insert_FINAL_8-30-06.pdf

and the fact the the Toshiba HD-XA2 player went up to using a 12 bit video DAC for component video output -- I wonder how much of the banding being reported is display related?

Using component video connection from a XA2 to a CRT based display would seem to give one 36 bit color display capability right now.
Are not "Deep Color" HDMI 1.3 LCD displays due to be available soon or even now?
http://www.tacp.toshiba.com/televisions/product.asp?model=52LX177It's available now, which are x.v.Color Bravia TVs.

Would it be feasible to implement xvYCC in future HDDVD/Bluray players movies? I know it isn't in the specs for either formats. If they could implement it into the spec im sure movies released up that would point would be unchanged but all movies and player from that point forward having them wouldn't seem that hard to do.

http://www.hdmi.org/pdf/HDMI_Insert_FINAL_8-30-06.pdf
It enables increased contrast ratio, and can represent many times more shades of gray between black and whiteYeah..well according to this forum greatest minds :)
bit depth and contrast ratio have zero correlation
The conjecture is usually followed by a painful dissertation utilizing a light bulb and ON/OFF switch analogy, suggesting the hypothetical has infinite contrast. Heres their mathematical proof ;)

1 / 0 = CR

The above is again proof positive the dividing by zero is a "weapon of math destruction"

Would it be feasible to implement xvYCC in future HDDVD/Bluray players movies? I know it isn't in the specs for either formats. If they could implement it into the spec im sure movies released up that would point would be unchanged but all movies and player from that point forward having them wouldn't seem that hard to do.
Yes, but if the format war has shown us anything is that people with current players aren't happy when new movies won't play :).

The only way this could work is if there was a backwards compatible stream and some kind of enhancement stream that newer players could take advantage of. There's been R&D around this, certainly, but it's far from trivial.

Ben -

Is there a 16 bit 4:4:4 version of VC-1 in the works to target the next generation of holographic disc players? ;)

- Tom

Is there a 16 bit 4:4:4 version of VC-1 in the works to target the next generation of holographic disc players? ;)
Well, for a DCT based codec like VC-1, it's pretty obvious how to extend it to support greater precision and bit depth.

Let's just say we'll probably have something ready by the time holographic storage is safely a mass-market technology :)...

Okay, I went ahead and made a much more detailed description of how I encoded the Elephant's Dream clip, this time including a normal download link (no .torrent client needed).

Now that I've got the hosting side of things figured out, It'll be easier for me to provide future sample clips around VC-1. I hope to have a good workflow soon to be able to mux HD DVD and BD compatible VC-1 streams into a .wmv file for demos specifically about HD optical disc.

Okay, I went ahead and made a much more detailed description of how I encoded the Elephant's Dream clip, this time including a normal download link (no .torrent client needed).Very nice. With Vista64 it started Media Player when I hit the Start Download Now link. So it played the clip and I was able to easily save it from Media Player when it finished( 850,666 kbyte size for 10 min 58 sec, 25 Mbps video). Looked very good up close on a 17" LCD. Notice it uses 5.1 ch 24 bit 48 kHz 429 kbps WMA audio. Nice bass. :)

Thanks! I forgot to even mention the audio - it's become almost boringly easy to deliver awsome audio with WMA Pro now. No special sauce there at all, unless you count using VBR (many forget, but if video is VBR, audio might as well be too).

As promised way back when, I finally posted up that 2 Mbps sample.

It's great for 2 Mbps, but there's a reason Xbox Live Marketplace uses a lot more bits for 720p!

http://on10.net/Blogs/benwagg/elephants-dream-720p--2-mbps/

As promised way back when, I finally posted up that 2 Mbps sample.

It's great for 2 Mbps, but there's a reason Xbox Live Marketplace uses a lot more bits for 720p!

http://on10.net/Blogs/benwagg/elephants-dream-720p--2-mbps/

Ben, the link appearing in your blog is broken.
Here's the correct link:
Elephant's Dream 720p 2M.wmv (http://download.microsoft.com/download/9/b/4/9b4535e2-f908-4785-97a5-ef360405e391/Elephant's Dream 720p 2M.wmv)

BTW, looks awesome for 2.0 Mbps. :)