AVS Forum banner
Status
Not open for further replies.
1 - 20 of 44 Posts

·
Registered
Joined
·
304 Posts
Discussion Starter · #1 ·
For the WM9 codec, is MS claiming HD resolutions at DVD bandwidth at cinema frame rates only (24fps) or are they including all HD standards?


Because our HD standard also includes 720p at 60fps, and I'm having a hard time believing that they could pull off the same magic at 60fps using the same "low" bandwidth.


Also if you do the "digital camera comparison" analogy, 1280x720 = .91 megapixels/frame. But 1920x1080= is just over 2 megapixels/frame.


So considering that WM is crunching less then half the pixels over "true" HD, the fact that it can do it with half the bandwidth of ATSC (incidentally the same bandwidth of current DVD's), well it doesn't seem that magical to me.
 

·
Banned
Joined
·
18,829 Posts
You are correct that when we say we need 5 to 7 Mbit/sec to produce HD content on DVD, we mean 24fps. But this is also true of DVD format and its 24fps MPEG-2 encoding (the output comes out at 60i or 60p but the source is usually 24fps). It would not be fair to ask us to produce 60 fps when the current DVD format does not use such a source.


Also note that I am not aware of a lot of 60 frames/second programming. Much of the (TV) programming out there is at 60 fields (as in 1080i), not 60 frames/sec. So roughly speaking, the frame rate is 30% more than 24 fps (you can imagine 60 fields being 30 frames progressive), not double.


Even if the frame rate were to be 60fps, there is so much more correlations between the frames that you would not need anything close to linear scaling to accommodate them. I estimate that you need 20 to 30% more bits to represent the additional frames.


But again, you need to compare the systems with the same frame rate. I do not think MPEG-2 at 60fps would look great at say, 5 Mbit/sec. For the same reason, 60fps 720p WMV9 would not look as good at 24fp at the same rate. You would need to boost the data rate higher to accommodate 60fps in MPEG-2 and do the same with WMV9.


As to 720p versus 1080p, again, the linear ratio can be deceiving. We can get superb quality 1080p content at 7 to 8 Mbit/sec. This compares to 6 to 7 Mbit/sec for 720p. So doubling the pixels does not result in doubling data rate due to higher efficiency of encoding higher resolution frames. But you do need a few more bits for comparable quality.


Amir

Microsoft
 

·
Registered
Joined
·
6,227 Posts
What approx processors are the 1080p sources going to need... > 3Ghz or not... I am playing chicken and egg as my current 2.4 will handle the 720p source and as yet I have no movie content to play... I will wait until I have an uber rez source on the horizon before upgrading...


In related thinking, what news if any can you tell us about WMP9 assiting GFX cards... Heard some rumors but nothing substantial... Who makes them ??? What kind of performace changes can we expect ??? Etc..


Lastly, what are you doing answering our questions ?? I expect you out there lobbying the industry... Hell if no one will listen walk up and down Hollywood boulevard with a sandwich board "The end is nigh... WMP9 is coming..." of something subtle to that effect :) !!!!


I wonder if many US based AVS'ers realize how much pent up (frustration) and demand there is among the (admitedly small but ohh so keen) Euro and 'other' populations that have HD capable HT's but no source...
 

·
Registered
Joined
·
304 Posts
Discussion Starter · #4 ·
Okay, but how would the quality compare to proposed blue-ray HD discs, assuming for the data rate is similar to D-Theater (MPEG2/26mbps).


The problem I forsee is that if the industry & public accepts WM/red-laser based HDTV, they won't notice how shortchanged they're getting by limiting themselves to the red-laser bandwidth.


Although there isn't much 60fps programming yet, it doesn't mean there wouldn't ever be. It's still a part of the ATSC standard, so the option is out there.


I'd just hate for a racing fan to buy a "Nascar 2009 Season Highlights" HD-DVD only to find that all the 720p/60 action has been dropped to 30fps in order to squeeze it within red laser bandwidth.
 

·
Registered
Joined
·
6,227 Posts
While I agree no one wants to settle on a weak format I want some content NOW.. Not in 2 or more years once they argue out more saftey nets for the content owners and other issues that feel more like stalling than anything else...


Secondly I dont think that WMP9 is going to be a STB kind of format due to the cost of the processing components... By the time the HD consortium are ready to put out content, if WMP9 has some following then they will have to make sure it stands head and shoulders above the competition...


I cant believe that the quality of sources are going to stagnate (look at the progress and rate of progress re VHS / LD / DVD / WMP9) so basically I see any movement to higher source as a posative... The next one has to beat it to have any hope... By wanting it NOW you gte the time pressure to evolve...


Perhaps I am a glass half full kind of guy though :) !!
 

·
Registered
Joined
·
6,227 Posts
Ohh yeah forgot to say... By the 2009 stage when we have blu-ray discs in PC's anyway.. Whats to stop WMP9 having 1080p or the 60fps or whatever is popular ?? Being a adpatable codec surely gives it more forward capability... It does not need to be stagnant...


By the time any programming is out there we should have enough storage and horse power to use it...


You know how MS could use a very different tangent to win a lot of naysayers... Use thier muscle and push to make a true high end DVD-A player and force the issue in a way M-Audio seem to have failed in... If they show high quality can be achieved (in a difficult subject like PC audio) they would get more trust and leeway in the industry with high end sources...


From what I can read between the lines... It looks like producing a SW DVD-A player is only a small part of the difficulty of getting one to market... Dealing with the content owners is what appears to be a Maciavellian wet dream / nightmare...
 

·
Banned
Joined
·
18,829 Posts
Our current spec for 1080p is a 3 Ghz P4 CPU or equiv AMD. But as I mentioned in another thread, the GPU needs to have a fast transfer method to its frame buffer or you will have a glitch or two. And of course, the encoding method makes a difference. We can encode clips that even a 3Ghz CPU will drop some frames on but we are careful not to do so.


BTW, the "official spec" for T2 calls for 3 Ghz CPU (we are hoping to tune it so that it works better but this is what the safe answer is). And no, this doesn't say it comes at 1080p :).


As for blu-ray, MPEG-2 at 24 Mbit/sec looks the same as WMV9 at 7.7 Mbit/sec based on tests conducted by DVD-Forum in the process of selecting a red-laser DVD codec. So you do not get better picture quality with blu-ray and MPEG-2. However, combine blu-ray and WMV and now we are talking. You can easily scale to 4Kx4K resolution @20+ Mbit/sec WMV9 or put a bunch of movies on one disk. You get to pick! Blu-ray is fine technology but putting a 10 year old video codec on it is like having a 2003 care with a 1993 engine in it!


Hardware acceleration of WMV in the graphics processor should be available later this year.


As to me answering questions here rather that spending time on the sidewalks in Hollywood, I think the former is more fun :)


As for STB processor requirement, WMV requires 2X the resources of MPEG-2 (but less than half of H.264). This is pretty low requirement given how long MPEG-2 decoder ICs have been around. Indeed, we have shown for a year now an equator chip that is able to decode WMV at 720p. Polaroid showed a prototype DVD player using the same chip at CES. The fact that we can do pure software decode of 1080p should also tell you about the careful design of WMV from MIPS (and memory) efficiency point of view.


Amir

Microsoft
 

·
Registered
Joined
·
206 Posts
It's fun now that companies start thinking about increasing the source medium and material quality.


I watched the liquid2, confidence and foreign affair trailer on MS website. But I found the result clearly lacked in comparison with D-VHS.


1. At what bitrate does the Ant-War syndrome in skies etc go away? I could clearly see dithering all over the place. The effect reminds me of my old dot-matrix printer.


2. One movie trailer showed a desert scene with a blue sky. The sky was however banded, ie. instead of a continous shift of color gradients, there was bands that got me thinking of DIVX artifacts.


3. The Red channel bled and was of lower resolution than the rest of the colors. Clearly visible on the red and white text of the confidence trailer.


4. How many different colors can the WMV9 codec display in one frame? 256?


/Best Regards,

Fredrik Rasmussen
 

·
Banned
Joined
·
18,829 Posts
Quote:
Originally posted by Fredrik Rasmussen
It's fun now that companies start thinking about increasing the source medium and material quality.


I watched the liquid2, confidence and foreign affair trailer on MS website. But I found the result clearly lacked in comparison with D-VHS.


1. At what bitrate does the Ant-War syndrome in skies etc go away? I could clearly see dithering all over the place. The effect reminds me of my old dot-matrix printer.
If you are referring to liquid, I explained this elsewhere. But briefly, the clip has a ton of film grain and preserving this was a key requirement for the codec for HD and d-cinema. If we can preserve "noise", we can preserve any other high-frequency information. We can pre-filter this and come up with a smooth image but this is not what the director/producer wanted. Put another way, there is no bitrate at which the "ant-war" go away because the higher the bitrate, the more clearly they are perserved! That same detail is in the 35 mm source (I know since we have the original D5 master).


I don't understand the comparison to D-VHS. I have a ton of D-VHS content and some look very good, others look very soft. The ones that look good are shot on 1080i cameras which provides a level of brilliance that is hard to duplicate with film stock and telecine transfer. But then again, that is not the look that everyone is looking for. In addition, 1080i is 2X more resolution than the 720p clips that you see on our site so there is bound to be an obvious quality difference. Most importantly, we have not put on a comparison test against D-VHS. It is impossible to compare these two systems with different clips and/or processes to arrive at the content -- especially when one is driven by a PC and the other by an external deck (i.e. they are not calibrated the same).


When the same process is used for both technologies, WMV9 compares quite well to D-VHS. For example, DVD Forum in the process of selecting the codec for red-laser HD DVD used an MPEG-2 24Mbit/sec (specfiically from D-VHS encoding) as a reference with the proposed codecs at 7.7 Mbit/sec, 1080p. WMV managed to match D-VHS in 2 clips and came within 95% of its performance on the balance of the clips (WMV9 was the only one to achieve this level of performance).


In the above case, sources were identical D5 tapes and the outputs were judged essentially indistinguishable from D-VHS on double-blind test using broadcast HD monitors ($20K each) and experienced viewers. We presented these results at NAB conference last week to hundreds of people and are available from other DVD forum members. So when I say we compare favorably to D-VHS, it based on solid tests that were designed to not favor MS in any way :).

Quote:
2. One movie trailer showed a desert scene with a blue sky. The sky was however banded, ie. instead of a continuous shift of color gradients, there was bands that got me thinking of DIVX artifacts.[/b]
Good observation. This may be the quantization noise that you get with all codecs. Next time there is slow atomic explosion (e.g. in dinosaurs) check out the smooth graduations on the DVD and you will see the same steppiness with MPEG-2 there. This kind of distortion is very hard to avoid once you go in and out of the compression domain (even when the codec is set to its highest quality mode).


Also note that this may be caused by your graphics card not being set to display 16M colors and/or a digital display with poor A/D converters.


Quote:
3. The Red channel bled and was of lower resolution than the rest of the colors. Clearly visible on the red and white text of the confidence trailer.


4. How many different colors can the WMV9 codec display in one frame? 256?
Of course not. Hopefully are not serious :). 256 colors would look like paint by numbers. WMV9 uses 4:2:0 as does MPEG-2 in HD profiles and below. This means the color is subsampled compared to luma in frequency domain. In dynamic range, each chroma sample has 8-bits. Converted to RGB, this will result in 16 million colors, again the same as MPEG-2 HD. 10-bit YUV is only employed in studios, and not in consumer delivery formats.


At to red channel being lower resolution, this is not a function of the codec. The codec, if you will, is color blind. Your display has far more to do with this than any other component. The next factor is the color correction applied after the telecine transfer. This is a mandatory step to go from film gamut to video. This is a creative process performed on a hardware DaVinci color corrector (read super expensive) that performs 16-bit per component color correction so I doubt that there is any resolution limit there. But the post house can take liberties with what they do here.


Amir

Microsoft
 

·
Registered
Joined
·
304 Posts
Discussion Starter · #11 ·
Obviously WM is biased torwards 24fps formats, since pretty much the whole film industry is based on it it's easier for software players to deal with them.


My only concern was for the preservation of ATSC formats across any future HD media standard.


ABC & ESPN have chosen 720p/60 as their choice for their sports coverage. And I'm skeptical with your assertion that WM 720p/60fps would only need 30% more databits then 24fps. For example an in-car camera of Formula-1 car would only need 9mbps @ 720p/60f ? .. and what would the CPU requirements be for smooth playback of such on a software player?


Also how does WM format come into play with existing (and future) firewire based standards- when such displays lack WM codecs? Is a codec crossconverter box possible?
 

·
Registered
Joined
·
178 Posts
Quote:
Obviously WM is biased torwards 24fps formats, since pretty much the whole film industry is based on it it's easier for software players to deal with them.
WM is not in anyway biased toward 24fps, it simply takes more bps to store more fps, mpeg-2 needs this as well, only more bps than WM.

Quote:
My only concern was for the preservation of ATSC formats across any future HD media standard.
Blue-Ray, is not preserving ATSC, neither is D-VHS, these are both storage medium not compression codecs, both mediums can store WM, or MPEG-2 or any stream of bits you like.

Quote:
ABC & ESPN have chosen 720p/60 as their choice for their sports coverage. And I'm skeptical with your assertion that WM 720p/60fps would only need 30% more data bits then 24fps. For example an in-car camera of Formula-1 car would only need 9mbps @ 720p/60f ? .. and what would the CPU requirements be for smooth playback of such on a software player?
Why is is hard to accept that a temporal compression algorithm scales non linearly with a higher frame rate? This is the bases of how mpeg and WM achieve such high compression compared to a 2d codec like MJPEG. Regardless of your skepticism, be assured mpeg-2 requires just as much more % bits for it to do 60fps compared to 24fps, and alot more than WM does ;).

Quote:
Also how does WM format come into play with existing (and future) firewire based standards- when such displays lack WM codecs? Is a codec crossconverter box possible?
This is a basic problem with building a codec into the spec for a display interconnection, with the current firewire standard you are stuck with mpeg2 everywhere even if something better comes around like mpeg4 or WM. This is why alot of people are debating firewire vs DVI, since DVI is raw uncomressed video. But theres nothing to say that firewire can't transport WM instead of mpeg2, you will just need a decoder chip in the destination device than can decode either. It is very similar to having a pre/pro with DD before DTS was available, if you want to listen to the DTS track you need a new pre/pro. With firewaire its completely possible to create a trancoder box to go from WM to MPEG2, how much that would cost is another matter entirely.


Justin
 

·
Registered
Joined
·
9,884 Posts
Quote:
ABC & ESPN have chosen 720p/60 as their choice for their sports coverage. And I'm skeptical with your assertion that WM 720p/60fps would only need 30% more data bits then 24fps. For example an in-car camera of Formula-1 car would only need 9mbps @ 720p/60f ? .. and what would the CPU requirements be for smooth playback of such on a software player?
It may be changing now but for all of last (calendar) year I am not aware of anything sent by ABC that actually was shot with 60p cameras and transmitted 60 unique frames / second. Most ABC material in my area is sent with duplicate frames, not even using the MPEG-2 repeat flags.


I'm also skeptical about the 30% increment. When ABC finally sent some real 60p material with this year's Superbowl I recorded it and did a couple quick tests of what it would take to re-encode into MPEG-4/Xvid, at 20, 30, and 60 FPS. I didn't have any conclusive results but it generally seemed that if I increased the frame rate by a certain percent the needed bit rate or file size would go up by about 3/4 of that.


This was done using a fixed quality (fixed quantizer=2) setting and then comparing the size of the files.


You would think that normal motion compensation would make it better than that but it's not the results I got. But again, this was with MPEG-4 so maybe WM9 is different. It would be easy enough to find out.


- Tom
 

·
Banned
Joined
·
18,829 Posts
The 30% number that I put out was just an average based on some tests I ran a while back. Your mileage may vary. You can have content the requires more than this if there is a ton of motion. But on a long clip, with a combination of slow and fast scenes, the average should be lot less than 3/4. Of course all bets are off if you are shooting basketball :).


Amir

Microsoft
 

·
Registered
Joined
·
9,884 Posts
Okay, that makes sense. And I should confess I chose a section containing both some high action football and a detailed pan over the crowd.


OT: For a couple years one of my main test disks for writing deinterlacing code was the NBA Jam DVD. This is a strange music video made from basketball scenes and is a mixture of video and telecined material. I got so I could hardly stand to watch NTSC bball, both for the association and the artifacts. ;)


- Tom
 

·
Registered
Joined
·
13,531 Posts
Quote:
In the above case, sources were identical D5 tapes and the outputs were judged essentially indistinguishable from D-VHS on double-blind test using broadcast HD monitors ($20K each) and experienced viewers. We presented these results at NAB conference last week to hundreds of people and are available from other DVD forum members. So when I say we compare favorably to D-VHS, it based on solid tests that were designed to not favor MS in any way
Can you tell us the size of the broadcast HD monitors used? And the viewing distance?


On the HDTV programming forum, we see some members with new 30" to 34" 16:9 televisions comment that they don't see much (if any) improvement over a quality DVD with the film-sourced content on ABC, CBS, and HBO. But then, these people aren't doing A/B comparisons of the same film, but just making general, first-impression statements. Obviously, quality difference one is able to discern will be a function of display size and viewing distance.
 

·
Registered
Joined
·
5,342 Posts
Quote:
In the above case, sources were identical D5 tapes and the outputs were judged essentially indistinguishable from D-VHS on double-blind test using broadcast HD monitors ($20K each) and experienced viewers
And that is where your quality assessment goes out the door.


When you run the same A-B comparison using CRT FP on a reasonably sized screen ( 8' to 10') width with a viewing distance of 8' to 10' and you can make the same statement then you have got a valid comparison.


All you have to do is look at the crap that passes for high quality DVD and the arguments from the studio that EE (or whatever you want to call the halos around objects) doesn't exist to know that the monitors used for QC in the telecine and compression houses don't do a decent job of allowing effective QC. Many of our HT's are far more revealing of the quality of the image.


Vern
 

·
Registered
Joined
·
2,541 Posts
One of the BIGGEST differences between MPEG2 and WMV/MPEG4 is that the newer codecs use a FAR greater distance between key frames. In MPEG2, you usually have around 2 key frames per second. In WMV/MPEG4 the key-frame distance can be as far as 30 seconds (usually limited to under 10 seconds). This distance between key frames is one of the key factors that improves compression (as key frames don't compress as well and less of them equels better compression).


This is where the MPEG4(DIVX/XVID) and WMV type artifacts come from. You see, in MPEG2, twice per second, these artifacts get reset, so it's percieved as grain/noise (which the human eye is used to), but with these new codecs, the artifacts are maintained, leading to what is percieved as a blurrier image. Post-Processing the artifacts makes the image slightly softer as well.


The solution? Higher bitrates... Less artifacts, less post-processing.


So even with MPEG4/WMV looking good at about 1/4th the bitrate of MPEG2, I think that more keyframes should be introduced (also to enable fast forward/rewind and lightining fast seeking) and the bitrate used at around 1/2 of MPEG2.


Just my 2c.
 

·
Registered
Joined
·
304 Posts
Discussion Starter · #19 ·
I know the difference between codecs, bitrate, & delivery systems. my concern is that if a MS red-laser becomes some sort of HDTV standard by market penetration; other formats will suffer.


I mean-- new formats can enter the market, but they are essentially thwarted by the presence of a dominate market, and thus become niche markets (see VHS vs laserdisc.).


I'm not particularly anti-MS, it's just hard for me to think of it as a good thing when one company can hold the keys to so many industries. Perhaps its just my "fear of an MS planet". ;)
 

·
Registered
Joined
·
9,884 Posts
Quote:
my concern is that if a MS red-laser becomes some sort of HDTV standard by market penetration; other formats will suffer.
It is a concern, but not a very large one to me. If you look at the various standards of TV's, audio & video tapes, discs, etc. the first one to market often was not the eventual standard.


And generally I don't think it is ever fair to say someone should not bring a desired product to market just because it may make it harder to compete for future products that aren't even finalized yet.


This is especially true when the body that claims to be setting the "standard" already effectively represents a competitive product (DVD's) and has a history of stalling for years bickering about copy protection issues. So it is no surprise that there is quite a PR campaign going that only their one standard should be allowed to compete.


- Tom
 
1 - 20 of 44 Posts
Status
Not open for further replies.
Top