AVS Forum banner
Status
Not open for further replies.
1 - 20 of 30 Posts

·
Registered
Joined
·
12 Posts
Discussion Starter · #1 ·
There is a comparison of 7 different video codecs over at doom9.org . They used 3 dvds as the source. They were Matix, Saving Private Ryan, Futurama Season 2. Windows Media Video 9 did very good, but wasn't the best according to them.


The videos are encoded at between 581k/s and 1016k/s. It would have been interesting if they had done some comparisons at higher bit rates.


Judge for yourselves. Complete with screen shots.
 

·
Registered
Joined
·
1,750 Posts
They start with this quote - "Then my usual disclaimer: Screenshots are not necessarily representative of the video quality." - then immediately go on to draw video quality conclusions based upon subtle subjective differences between screenshots.


Something like PSNR4AVI that objectively compares and measures pixel by pixel differences of entire clips is not perfect, but is easy to set up and might be better suited to the task.

http://www.vsofts.com/codec/codec_psnr.html


Joe
 

·
Registered
Joined
·
219 Posts
What would be the best codec for encoding satelite tv, at full resolution and no audio comprecion. ??


I know some use Divx or Xvid but it's not satisfying to me. I would like it to look just as it did before encoding. Is that posible?


Perhaps there might be a guide somewhear out there??


Regards


Skipper
 

·
Registered
Joined
·
3,496 Posts
Gotta laugh when the are shooting for the best Codec and then limit themselves to such low standard codecs and DIVX, and Real. Nothing wrong with them but the best that can be expected from them is TV quality.
 

·
Registered
Joined
·
832 Posts
We participated in previous shootouts with our VP4 codec, (insert shameless plug here: please visit www.on2.com for more info!) but didn't participate in the newer one primarily because of the reasons you state. The tests and conclusions were a bit too unscientific, especially when you consider the differences between playback on a PC and other devices (which is where we focus).


Although non-ideal, we favor using an objective, hard metric (PSNR for now), and you'll see that as we launch VP6!


Eric.
 

·
Registered
Joined
·
832 Posts
Codec = Compressor/decompressor. Video data in it's raw form is just too darn big, and needs to be compressed (a whole lot) to get it to fit on a disk (or other storage medium), or through the pipes across networks and/or to your home.


A 2 hour raw video at NTSC resolution would take up on the order of 150GB of data, and a connection speed in excess of 25MBytes/sec.


Eric.
 

·
Registered
Joined
·
9,884 Posts
Doom9 has put in a lot of work into these ungoing shootouts and has gotten quite good at them. I tend to give them a fairly high amount of credibility, even if they are somewhat subjective.


I also believe that both the metrics like PSNR and also comparison shots have to be taken into account. So the side by side shots at the same bit rate and the commentary are quite valuable, even if they are not the whole picture.


That said, realize these are targeted mostly at 1 & 2 CD rips of DVD's, at probably lower bit rates than we would use here for HTPC purposes. This becomes even more true as we all get HDTV's and DVD burners.


Even with that, though, it seemed to me that Xvid and WME9 informally seemed to come out mostly the winners. And this is consistant with my own experience and reading recently. (what a surprise ;) )


It is too bad though that H264 didn't make the cut this time around. Hopefully it will be ready by the next time.


- Tom
 

·
Registered
Joined
·
262 Posts
Quote:
They start with this quote - "Then my usual disclaimer: Screenshots are not necessarily representative of the video quality." - then immediately go on to draw video quality conclusions based upon subtle subjective differences between screenshots.
Um no they don't. If you read the article you'll notice that the author is not basing his reviews on the screenshots at all. Many of his comments start with "you can't see it in the screenshot, but", or include "the screenshot does not do this justice" (especially the Futurama comparision where the author basically says the screenshots are nearly useless). I thought it was pretty clear that the author was reviewing the final encoded video, and that the screenshots were just to give a rough comparision (i.e. SBC is detailed but blocky, 3ivx looks like crap, etc.).
 

·
Registered
Joined
·
12 Posts
Discussion Starter · #11 ·
Also take note that they are compressing DVD, which is already compressed, even further. Some of the codecs might not be optimized for this. I would like to see a comparison of previously uncompressed raw video.
 

·
Registered
Joined
·
602 Posts
I'm glad someone brought this up. I read that "comparison" and came away thinking that it was a total waste of my time to read. Why in the hell are they comparing the encoders in such a stiffled limited way? Packing a 2 hour movie onto a CD and having the nerve to try and compare it to DVD quality is obsurd. I would have taken the article more seriously if the experiments were of the encoders full potential not they're barest stripped down potential.


Give me a serious comparison of the encoders using real bitrates that CAN be DVD quality, not lowball bitrates to fit on CD's so cheap ass college kids can watch movies on they're computer screens.


In case no one noticed, DIVX is providing DVD quality "Home Theater" and High Def quality profiles to encode with. This means there must be an interest to encode in better than Compact Disk space.
 

·
Registered
Joined
·
684 Posts
This test is interesting when you consider the upcoming WMP9 HD disc of T2, and other upcoming HD movies. Compressing an HD-res movie onto a DVD is like putting an SD-res movie onto a CD.


In this light, it is interesting to see how WMP9 compression compares to other methods. It will be even more interesting to compare WMP9 HD on DVD to the MPEG2 blue light DVDs (when they become available).
 

·
Registered
Joined
·
720 Posts
Shawn:


I would go one step further than your comments and say that future focussed interest is more in putting a 2 hour High Definition movie onto a DVD disc, as opposed to trying to put a 2 hour DVD movie onto a CD at VHS quality!


To do this testing correctly, they would need a clean uncompressed master 1920x1080p source video clip... not sure where you get your hands on one of these though? ;)


Then they could compare results of compressing this clean HD source down to the 5-8Mbps required to put HD content on an existing red laser DVD-5 or DVD-9 disc.


This is the future, where we want to understand which codec performs the best!


Greg
 

·
Registered
Joined
·
376 Posts
You guys don't understand. In avsforum the most interest is in compressing HD sources to dvd disks. The doom9 forum, however, is about getting the best possible quality when archiving SD to cd size (typically from dvd disks).


This is done primarily for PIRACY....although no one there will admit it. Do a search on emule, edonkey, or kazaa, or whatever and you will see thousands of cd-size rips of movies/television. 700megs is a reasonable download when you have broadband, and that is the current standard for releasing rips over the net.


Most of them couldn't care less about encoding hd content because it would be far to large to release anyway.
 

·
Registered
Joined
·
1,750 Posts
"To do this testing correctly, they would need a clean uncompressed master 1920x1080p source video clip... not sure where you get your hands on one of these though? "


You can get pretty close with AVIsynth and the "stack" filter. It will take any 9 high quality SD 640x360 clips (or 1 clip used 9 times) and tile them into an uncompressed 1920x1080 AVI which can then be used as source material for HDTV codec comparisons. Start with low motion video for the highest detail, then you can decimate frames to check the effect of rapid vs low motion encoding, etc.


Joe
 

·
Registered
Joined
·
9,884 Posts
And you can also get pretty close by oversampling. Start with a good 1080i capture, filter it some, and downsample to make a good 720p clip out of it.


- Tom
 

·
Registered
Joined
·
130 Posts
Quote:
Originally posted by jamoka
They start with this quote - "Then my usual disclaimer: Screenshots are not necessarily representative of the video quality." - then immediately go on to draw video quality conclusions based upon subtle subjective differences between screenshots.
Do you honestly think they make the comments based on quality from the JPG versions they show on the page?


They make them based on the original screenshots they take, then they compress them so they don't waste 10 megs of bandwidth a page just on PNG or BMP pictures.
 

·
Registered
Joined
·
850 Posts
jamoka:

I think the point was that regardless of what is used as the source, most of us only get 'compressed' source at this point.


Even getting the 1080i broadcast from HBO/Showtime/CBS, as good as they are, it's already MPEG-2 compressed (upto 20 Mbps).


The only non-compressed SD signal that I can think of would be ANALOG OTA SD signal (preferably from BUDs, goes through lesser analog degradation). Or, make your own - use a regular VHS VCR analog camcorder to record stuff (babies/parties/wedding/etc.). Then feed it into the capture card, capture at full resolution. DV-cams compress to MPEG-2, although at up to 16 Mbps?
 

·
Registered
Joined
·
12 Posts
Even analog OTA-SD via BUD is probably being fed from a compressed source - HDCAM if you're lucky, and maybe some kind of god-knows-what edit suite internal format if you're not...


Starting from a compressed source is perfectly valid IMO, since that is what most "real world" transcoding we consumers will be doing anyway.


MiniDV works out to about 3.5-ish MB/second, call it 28mbit, close enough. MiniDV supposedly has limited red bandwidth. Looks pretty good to me .


In my own testing, I was finding that at bitrates to transcode a DVD to 2-cd size (1200-2000-ish kbps) wm9 would do distracting things to background textures like flashing between sharp and smeared, while divx/xvid would just leave it smeared. Personally I find the constant smear less distracting than textures crawling around on supposedly static surfaces in the background.
 
1 - 20 of 30 Posts
Status
Not open for further replies.
Top