comparison of the different CODECS

[lilhobo]

has anyone compared the codecs in terms of quality speed and suport?

is there a site you can point me to for a ratings out of ten :D

especially divx and wmv...since its seem that real gets a fan base here due to support anwyas :D

[KpeX]

http://www.doom9.org/codec-comparisons.htm - best on the web.

Remember rule one. hth,

[Sirber]

No offence, but d9 comparison is outdated :)

[SeeMoreDigital]

Originally posted by Sirber
No offence, but d9 comparison is outdated :) Agreed. However, after readind a recent post made in the DivX 5 section I feel that Doom9 will be posting new test results very soon.

Cheers

[Hiro2k]

Look for the threads here in the New A/V forum for up to date comparison on all the newest and older codecs. They are really helpfull when your looking at encoding a specific type of video (anime ...etc). Though you must note that those become less usefull as time goes by because the codecs that are still chaning always are getting better.

[Ramirez]

Besides you can make your own mini-comparison (http://storm.wronger.com/570k/index.html) ;)

[Valky]

@Ramirez: nice comparison. And just what I expected. R9 is better than the rest almost in every frame, but not on those with explosions. It doesn't never create macroblocks, but blurs pretty much on those scenes.

But there can be seen, that from my opinion, it is the best choice in every other scene. It doesn't blur anymore like it used to and is at least as sharp as xvid on low-action scenes these days.

Windows media blurs too much. Everyone can see that.

[Sirber]

RV9 is a little more sharp than XviD and a lot more than WMV9. WMV9 is more blury than RV9, and less detailled than XviD. XviD produce macroblocks and muskito noise.

IMHO RV9 produce the best quality for this clip.

P.S. Don't tell me that PP will remove those artefacts, coz I know PP removes also details :)

[edit]

RV9 and WMV9 uses in-loop filtering at encoding time. I wonder how XviD will look with in-loop...

[Tommy Carrot]

Originally posted by Sirber


RV9 and WMV9 uses in-loop filtering at encoding time.

I know this is slightly off-topic, but WMV9 doesn't use in-loop filtering. ;)

[Sirber]

From the specs it does :p

[KpeX]

Nice work on the comparison. Xvid definitely doesn't do well at that kind of bitrate, but it still looks like the sharpest (sometimes more blocky) to me out of all three. However this is one of the first tests where i've easily been able to discern between RV9 and WMV9 so perhaps Real is improving...but I still watch my movies in motion not frame by frame :rolleyes: so that makes it tough to judge from that.

I'd agree with Sirber that filtering is having an effect on RV9 & WMV9, as usual.

[BoNz1]

Originally posted by SeeMoreDigital
Agreed. However, after readind a recent post made in the DivX 5 section I feel that Doom9 will be posting new test results very soon.

Cheers

I don't agree, what has really changed since that was done? Not much really, IMO and certainly not much to warrant another comparison. Sure, DivX 5, RV9, and XviD have made some improvements but I can't see why we would need another comparison. All the codecs are quite close as far as quality right now anyway much as they were when that test was done, I think you would find that this is still the case. And I never got the impression that there was going to be another one anytime soon from what doom9 said.

[Sirber]

We should wait for H264 to do a new one... IMO :)

[BoNz1]

My thoughts exactly, ;)

[Riley]

How feasible would it be to create a program, similar to what is on Rameriz's page, that would allow you to switch between various encodes of a clip while they play? That would allow you to compare both still frames and motion clips.

I'm no programmer, but I think it would be an incredibly useful tool if someone could create it. Even if it would only work offline.


Anyone care to comment?

[SeeMoreDigital]

Originally posted by BoNz1
I don't agree, what has really changed since that was done? Not much really, IMO and certainly not much to warrant another comparison. Sure, DivX 5, RV9, and XviD have made some improvements but I can't see why we would need another comparison. All the codecs are quite close as far as quality right now anyway much as they were when that test was done, I think you would find that this is still the case. And I never got the impression that there was going to be another one anytime soon from what doom9 said. Well we all have our opinions BoNz1.

Personally I think there should be another series of tests. And in my opinion they should all be at bitrates below 950kbps. As it's only when you encode at around twice this bitrate do the codecs begin to look the same!

I also think that any new tests should involve more PAL sources as these are harder to encode than their NTSC equivalent!

And if you don't think that's the case. Why not try this? First encode an NTSC DVD with an aspect ratio of 4:3 which has an image that occupies all the pixels. Simple so far, is'nt it? As an NTSC user you can resize the output to 640x480 and as such maintain the 4:3 aspect ratio and all the vertical pixel resolution - And you even get to save on the total number of output pixels to boot!

Now try and do the same with an equivalent PAL DVD which also has an aspect ratio of 4:3. As an NTSC user, how would you resize the image
without loosing any of the vertical pixel resolution? Encode to 768x576 maybe! Well if you did that you would end up with a pixel width greater than (both a PAL or NTSC) anamorphic DVD. And finish with more pixels than you started with!

Some of us DivX, XviD, RV9 etc users have different requirements. Some of us want to play back our encodes via hardware. Some of us like to keep the output resolution of our encodes the same as the input resolution. Because some of us like to view our encodes on large monitors of even projectors!

But, going back to the start of my post (I'm now speaking to those people who have been bothered to read all this babel).

Have we forgotten what the primary goal of all these 'non Mpeg2' codecs is? Which is allowing us to store what would have been large sized video files using higher bitrates, into smaller sized video files using lower bitrates. And all while keeping as much detail of the original source as possible!

While I admire the work that M$ has done creating high definition 1920x1080 (1440x816 actual) WMV9 encodes. They are hardly at low bitrates, infact they are higher than the Mpeg2/vob version.

Surley from a purley promotional aspect it would have been better to have created the high def encode the same size or less than the vob!

After all. Is'nt that what all these 'compression' codec's are about?

Cheers

[Gaia]

Use whatever codec suites best for your needs. Why some people are so obsessed about these tests?

Does it really matter what codec is the best?

How you can judge what codec is best?

It's impossible to do perfect test. Different sources, different display devices, speed, filtering, resolutions etc. Some codecs are free, some open source, some partly free. If you're developer, you can do special build for every XviD encode. That's one advantage.
In some rare cases SBC encoding is still best or atleast easiest way to encode few movies.

Example if some codec is slighly better in quality but needs lot more cpu power than others, is that codec really best in real life use?

If you know well how some codec works i am sure you can get better results than others. Better results than Doom9. You can do hunreds of tests and still some people will disagree. Testing Audio codecs is a lot easier because you can do those blind tests. Lots of people take part in these tests.

Video codec tests are usually made by one person. Sure most of the testers publish what settings and sripts they used so you can do it yourself but still...

I don't think people will start using different codec no matter what some test says. Just use whatever looks best for your eyes. Speed is important too if you don't have latest Pentium or AMD processor.

Just my thougts....No need to flame if you don't agree :)

[Doom9]

you can also make blind tests with video. However, the sheer size makes it a lot harder to allow a lot of people to participate. Maybe, if I ran a big shop that paid well I could try to negotiate something with content providers and get hosting to offer test files for download, but as things are now I don't see that happening.

As for settings, I think it's fair if the codec developers are informed about the source to be encoded, so they have the option of actually try and find parameters that show their product in the best light. If they don't, that is their fault. In my last test, not every company I contacted participated, but after the test got so much attention I got a lot more feedback, so the next test should be conducted using codec engineer approved settings for each codec included, leaving no room for errors on my part.

[SeeMoreDigital]

In my last test, not every company I contacted participated, but after the test got so much attention I got a lot more feedback, so the next test should be conducted using codec engineer approved settings for each codec included, leaving no room for errors on my part. Great news. Glad to hear it!

We've managed to do some blind tests at work using low bitrate (625-675kbps) encodes and showing them on a 42" plasma. Curiously most people prefer WMV9 with RV9 a close second! Goldmember and StarWars 2 DVD's were used as a source. We quite often use these encodes to set up the A/V equipment on exhibition stands. So quite a few people are able to comment. Not just those who are in the know!

Cheers

[General Lee D. Mented]

I'd say my codec is the best, but I dunno cuz it doesn't fully work yet. :P

[PatchWorKs]

Wanna see Rududu (http://rududu.ifrance.com/rududu/) in codec comparsion !

[superdump]

Originally posted by General Lee D. Mented
I'd say my codec is the best, but I dunno cuz it doesn't fully work yet. :P

LOL. :D It will soon enough GLDM.

[deXtoRious]

It's extremely hard to use Rududu in a comparison, because it's bitrate mode is so buggy, that I got a 25mb file instead of 10mb!

[Tommy Carrot]

Originally posted by deXtoRious
It's extremely hard to use Rududu in a comparison, because it's bitrate mode is so buggy, that I got a 25mb file instead of 10mb!

For reasonable bitrates, it works. I think you've gone under that. :p

[CruNcher]

i think their is no need for a H.264 test as the official tests of the whole idear not only some codec implementations will be starting soon and the results are expected to arive @ the end of this year part off the test draft are mathematical and blind tests with Normal and High Definition Content vs Mpeg4.

[deXtoRious]

For reasonable bitrates, it works. I think you've gone under that

Is 550kbps not a reasonable bitrate? :D
Anyway, I'll try 850kbps next.

[Sirber]

550 is is high for animes, and medium/low for DVD. It depends on the source.

[Dams]

and resolution .. ;)

[Sirber]

and skills? :D

[KpeX]

..and preference of detail and sharpness :rolleyes: 550 kbps....is too low for any DVD IMO

[Dams]

Originally posted by Sirber
and skills? :D

is it for me ? ;)

[Dams]

Originally posted by KpeX
..and preference of detail and sharpness :rolleyes: 550 kbps....is too low for any DVD IMO

+1 !

[SeeMoreDigital]

Hi,

It looks like I've come into this discussion quite late!

Personally I have not tried the Rududu codec yet but I'm willing to give anything a go!

The last time I followed Sirber lead, he turned me into an RV9 nut, so where this could go is anybodies guess!

My question is. As I personally prefer to encode at the same resolution as the source (ie 720x480 or 720x576) will this codec generate good looking encodes at bitrates between say 650 - 900kbps?

Thanks

[Lefungus]

Rududu has potential, but isn't quite up to current codecs. It generates too much color artifacts for that

[deXtoRious]

Sirber & Dams

I was asking if 550kbps was not enough for Rududu! :D

SeeMoreDigital

No, it won't. It doesn't work well until about 1200 kbps, AFAIK.

[SeeMoreDigital]

Bummer!

[Dams]

LOL :D

[Sirber]

:D

[CruNcher]

No, it won't. It doesn't work well until about 1200 kbps, AFAIK.


hmm not really true for me it looks @ low bitrates like 550 alot like RV9 but encoding time is alot really alot faster then RV9 this codec could beat RV9 with ease useing a good RC :)

[Sirber]

Nice. Is wavelet that powerfull?

[lilhobo]

all tests should have
1. speed.
2. low bitrates.
3. file size.
4. image quality


i dont see that doom9 tests covered all 4 aspects.

[Neo Neko]

Originally posted by SeeMoreDigital

I also think that any new tests should involve more PAL sources as these are harder to encode than their NTSC equivalent!

And if you don't think that's the case. Why not try this? First encode an NTSC DVD with an aspect ratio of 4:3 which has an image that occupies all the pixels. Simple so far, is'nt it? As an NTSC user you can resize the output to 640x480 and as such maintain the 4:3 aspect ratio and all the vertical pixel resolution - And you even get to save on the total number of output pixels to boot!

Now try and do the same with an equivalent PAL DVD which also has an aspect ratio of 4:3. As an NTSC user, how would you resize the image
without loosing any of the vertical pixel resolution? Encode to 768x576 maybe! Well if you did that you would end up with a pixel width greater than (both a PAL or NTSC) anamorphic DVD. And finish with more pixels than you started with!


That is simply because you are wasting CPU cycles and datarate to boot. PAL harder to encode? :rolleyes: The extra bit of vertical resolution on digital PAL sources is useless. It makes no easily percievable visual difference over NTSC. If you used a good resize kernel. Lanczos3 for instance you could enlarge the NTSC source to 768x576 and I have serious doubts that you would be able to easily spot any difference between the incorrectly resized PAL source and the NTSC source sized to match. PAL is every bit as easy to encode as NTSC. If you have to resize to a corect aspect ratio use 640x480 same as NTSC. Simple no? It is only as hard as you make it on yourself.


Originally posted by SeeMoreDigital

Surley from a purley promotional aspect it would have been better to have created the high def encode the same size or less than the vob!


High def gets its name from the higher definition that higher resolutions provide it.(much higher than the miniscule difference between PAL and NTSC) So to encode it to the same size or less than the vob would mean it was no longer high def.

Originally posted by SeeMoreDigital

After all. Is'nt that what all these 'compression' codec's are about?
Cheers

No they are about what they are designed for. No more; no less. They can all be designed to handle different things and behave differently.

[SeeMoreDigital]

Originally posted by Neo Neko
Lanczos3 for instance you could enlarge the NTSC source to 768x576 and I have serious doubts that you would be able to easily spot any difference between the incorrectly resized PAL source and the NTSC source sized to match.

I think you miss the point slightly. I'm not talking about resising an NTSC image to 768x576 as I would end up with even more pixels than I started with... When you generate an NTSC 4:3 resized encode from a DVD source, you will always end up with less pixels. When you do the same with PAL 4:3 DVD source you will always end up with more pixels (unless you encode using the anamorphic method). But that's not the point I'm trying to make!

Originally posted by Neo Neko
High def gets its name from the higher definition that higher resolutions provide it.(much higher than the miniscule difference between PAL and NTSC) So to encode it to the same size or less than the vob would mean it was no longer high def.

Not entirely true!
The high-def WMV9 version of Terminator2 could have been encoded at a lower bitrate so the file size could have matched the regular DVD. And as a quick test I re-encoded the T2 WMV9 trailer from off the M$ website to the same bitrate as the regular DVD. It's still high-def! It still looks superb! And let's not forget it would have looked even better if it had not have been a re-encode!

Originally posted by Neo Neko
No they are about what they are designed for. No more; no less. They can all be designed to handle different things and behave differently.

Agreed. What they were designed for is one thing. What they are used for is another.

And at the moment, most are used for compressing! How many DVD's have you encoded to say DivX, XviD, WMV9, RV9 etc at the same bitrate and/or file size of the original source DVD....? None I would presume!

Cheers