New mpeg-4 codec comparison on web!!

[Valky]

Hope you didn't miss this one. Unfortunately there isn't RealMedia or WM9 which would have been nice to see.


http://www.techimo.com/articles/index.pl?photo=33


One thing that I surely noticed was how good job that VP3-codec did on those low bitrate-shots.

They also found negative things on xvid-codec: Unlicenced... :)

[^^-+I4004+-^^]

be warned;
3.11 only vdub test
(no DRF tweaks of any kind,crispness100,
used fast motion codec for 311 and probably
for 320 too...)
test video 100frames long only
dial-up user be ware;
big as hell........
(IE got tired and at some point
didn't load no more pics,now trying my luck
with "kmeleon" browser)

but interesting anyhow........

[Doobie]

It's too bad someone went to all that work without very helpful results.

Grouping pictures by same bitrate is easier to compare than grouping by codec.

Very low and very high bitrate comparisons are not helpful to most of us.

Some important codecs were not compared while some unimportant ones were.

And, yeah, the clip is too short.

I'm still waiting for Doom :-)

[affter333]

Thanks,

much beter than Doom9's amateurish codec comparision

[Doom9]

@doobie: don't worry, it will come. But I do need to lay some groundwork first and the overhead in running a site and forum won't just stop from one day to the other (neither does my day job :/

@affter333: What is so amateurish about my comparison?

[affter333]

Originally posted by affter333
Thanks, much beter than Doom9's amateurish codec comparision
LOL I have violated forum rules because I think Doom9's
Codec Comparision is Amateurish.

I recieved the mail below:

MAIL FROM : Doom9's Forum Mailer <kaizen@doom9.org>

>> We have discovered that you have violated forum rule #2 in
>> http://forum.doom9.org/showthread.php?postid=294327
>> If you violate the forum rules 2 more times you will be
>> suspended for 30 days.


There's are reasons why not so many want to participante in here.

Moderators here are Shit Holes, I have seen many times they bash
people with harsh words becuase they asked the wrong Q,
or say things that they think are stupid or incorrect.

BUT you can not even remotely suggest that those Stuck-Ups
are anything but perfect.

Keep up the good work shitholes, and don't
forget to apply forum rules on your mothers. LOL



.

[Valky]

Originally posted by affter333
Thanks,

much beter than Doom9's amateurish codec comparision


I know right away when this post came up and I saw this guy's amount of post, that know we are dealing with forum-terrorist :)

Some unhappy user who just changed his profile, regged again and doesn't even appreciate other people's hard work..But one thing I dont get it..why bother to even make such insulting posts and all that effort?

You can't please everyone..I have seen that so many times now :)

[Doom9]

hmm.. it wasn't me. I wanted to strike first but thought "let's give the guy a break, maybe he has something useful to contribute". But now it looks like this isn't the case.

@affter333: You have 24h hours to explain why my comparison is amateurish, if you make a good argument I might consider removing that rule2 strike (but you have the rule4 coming for sure) which wouldn't automatically lead to suspension. I suggest you use your time wisely.

[Ookami]

@affter333

I have striked you. And you would get another strike 4 if it wouldn't would boot you out (I have never suspended anyone, and I don't want to suspend you). I think that a rule 2 is pretty cheap for your insult, it's not an argument, as you don't put any infos/proofs in it. I have yet to see a better codec comparison that Doom9's, I have seens dozens of them (probably all of those I can understand and that's quite a few).

If the moderators and this board sucks so hard then why are you here? Go somewhere else if you don't like it. Is it so hard?

BTW, you're incorrect and, you know. :D

@Valky

Noooo! I never thought of that ;) .

@Doom9

He insulted me, and I can go over it (rule4), but insult you in my forum will not be allowed. Especially when I see that 99,9% of the comparisons out there are really amateurish, your certainly isn't!

You know that I'm one of those persons who don't like to strike people, but sometimes it has to be done.

Cheers,

Mijo.

[NeVeRLiFt]

Only a Mpeg2 lover or a very slow thinking person would dare say such things....

I may not get along with some but I love Doom9 and this website ;)
Doom9 your comparison is second to none and the same goes for your guides! You have my respect and gratitude!!!

Gratitude is an attitude: a personal quality that molds us
and shapes our lives – not just something we do or say.


Mpeg4Life

[int 21h]

We're nearly to the point of 'Mpeg-4 is Mpeg-4 is Mpeg-4' and since the current rush seems to be heading into appliances (i.e. getting your product to play in a KiSS or similar), I can't really imagine moving away from this trend anytime soon.

Codec comparisons are interesting, but for the most part, unless you're a sheep, you're going to make your own conclusions. Not too mention, different settings on different targets will produce different results, and its unreasonable to expect someone to test every variation of every setting, that being said, a codec comparison can give you a good idea where to start looking.

Of course, this is only one small part of someone choosing a medium to store their content. If someone has a dvd burner and a standalone player that happily plays the media from that burner, more than likely they are going to select an Mpeg-2 solution (of course there will be exceptions...), while someone with a KiSS player (or similar) will be more likely to select an Mpeg-4 variant.

Unfortunately, this isn't even that interesting of a subject to argue about because of the immense amount of inhereit simularities in any software that implements Mpeg-4 concepts, instead, I would suggest arguing about containers and their merits. Then after we get the various factions with their zealots in on this.. the fun will really begin. ;)

P.S. - I believe that the linked comparison suffers from 'amateurish' content, not too mention its lack of HTML design knowledge. (Tables are so 1995! Not too mention the irresponsibility of embedding ~5MB of images in the document) Not too mention there are some blatant errors. For example, it is listed that HuffYUV is a lossless codec, however, Mpeg-2 is stored as YV12, and the conversion between YV12 and YUV is not lossless. Not too mention the overall premise between the comparison is suspect. For example, 3 of their codecs share the exact same codebase with only minor differences in constants within the code (DivX 3.11, DivX 3.2, Mpeg4v3), 2 other codecs also share nearly the exact same codebase (XviD and RealMagic). This is ignoring other issues, such as the fact that the original VKI hack for MS Mpeg4 (v2 and v3) (simply edited that low bitrate constant (460kbps) to a higher bitrate to make it apply variable keyframes to every frame) had huge issues, such as multiple consecutive keyframes in scenes with panning, etc. Plus sentences like this reveal the true knowledge level of an author:
...but this format fails to deliver acceptable image quality at low bitrates due to extensive macroblocking...
Macroblocking.. does he mean the process of subdivision of a picture during encoding or does he mean pixelation :confused:

However, that's just my two cents (maybe 2.5).

[NeVeRLiFt]

Pixelation and microblocks = the samething to me.... but a macroblock??? is that some kind of new macrovision :D

[PowerMacG4]

@Neverlift

I hope you are joking...

[^^-+I4004+-^^]

>Codec comparisons are interesting, but for the most part, unless you're a sheep, you're going to make your own conclusions.


hmmm........
no,if you're NOT a sheep you're gonna make your
OWN codec comparisons!

for example;should i really wait for d9's test every
few months (or years)?
what will his test tell me?
how do codec's act in event of dvd-rip
(not interesting at all,as i only do
tv/vhs)
how does "saving private ryan" look on one
cdr?
( not interesting too....for higher quality
i'll go for 2cdr's and hi-res....or more if need be...)
does rm9 still have flickering blocks (yes it does...
"real" rarely tweaks something after it's been released)
does wm9 have blocks and mosquiots as every mpeg4
(yes it does....this is just another tweak of ms mpeg4,
with hi-definition support )

now,don't get this wrong,it's my opinion..if i need
codec comparison i make it..i respect my needs
(my sources,my hardware will be used to play it,etc.)
and i do my codec comparisons.........i'm not biased towards
anything but use what gives me the best image

>Not too mention there are some blatant errors. For example, it is listed that HuffYUV is a lossless codec, however, Mpeg-2 is stored as YV12, and the conversion between YV12 and YUV is not lossless.

this is not so important,is it?
color conversions don't produce blocking..or?
perhaps it's not dvd source at all.....
(or you recognized scenes from some movie?heh)

>only minor differences in constants within the code (DivX 3.11, DivX 3.2, Mpeg4v3), 2 other codecs also share nearly the exact same codebase (XviD and RealMagic).

you might go much further with this,right?mention
"mumosys",history of ms mpeg4v3 and divx (and divx networks and project mayo etc.)
but is it really like that?
they are all different codecs after all and give different results...
(otherwise no one would be testing them if they are completely the
same...)
change one bit (or few) and you have another codec....
311 and VKI for example have exact the same dll,only
VKI kicks in not jsut below 460kbit,but works all the time
(this is regulated by few bits inside the codec..just a switch
so to speak)

>This is ignoring other issues, such as the fact that the original VKI hack for MS Mpeg4 (v2 and v3) (simply edited that low bitrate constant (460kbps) to a higher bitrate to make it apply variable keyframes to every frame) had huge issues, such as multiple consecutive keyframes in scenes with panning,

is this really an "issue"?
if i had space i would make my complete movies as KF's...
or better,i would burn my mjpeg directly to blu-ray disc....
there cannot be "too many" KF's,but there can be too few....
(all that thing with jerky playback on too many KF's..well,i didn't
saw it...ever.....perhaps it's just ffdshow being one mean
decoder....)
KF's are not some evil..they make mpeg4 work after all....
delta cannot be as good as KF on himotion scene...that's for sure...
sometmes MV's are not enough.....VKI is not that bad of a trick.....
although direct limiting of DRF's is even better....much better...

>Macroblocking.. does he mean the process of subdivision of a picture during encoding or does he mean pixelation

he means exactly that; macroblocking...you should know what "macroblock" is by now.....
so macroblocking means macroblock that's not placed correctly but
has visible MV error........
not so problematic term.......

but what is "pixelation"?
mpeg4 has mosquitos and blocking (these tend to be 16x16 pixel blocks)
pixelation would then be if i have 160x120 and blow it to full screen(?) how does that relate to the mpeg4 codec test....(?)

i mean,come on..let us not invent mistakes if someone else dared to make mpeg4 test (before d9 did).......

the test itself is not so perfect,but it has some highpoints;
for example it's good to include different bitrates
(sometimes i go for 900kbit,sometimes i go for 1800...it's good to
see samples for both...)
it's bad that i have to save images separately an dthen load to
image viewer for close examination,but any html test will
be like that......
it's too big too,but the results are pretty visible.....
i mean,give guy a credit..it's not small job......
also...to short clip?
well,yes,but if codec perform good in one-pass,it'll get better
in two pass.......
if it's sh*t in one pass ,then, well........
(for example "On2 VP3 3.2.5.0 " blurs too much on every bitrate...
a while ago when i tested it it reminded me of indeo...not to go into details,this simillarity has it's reasons....)

test images more or less confirm my experiences with different mpeg4 codec's.........

@d9
why get so excited about "affter333"?
he likes more pix ("BUT you can not even remotely suggest that those Stuck-Ups are anything but perfect")
i love few words thrown in too...

this comparison is decent one-pass comparison
(2-pass on 100 frames....yeah,right....)
and you make two-pass comparisons?
there's no match between "http://www.techimo.com/"
and "http://www.doom9.org/",if one is making 100 frames and another complete movie........
[and i'm not saying this to get into your as*.....i could post
100 questions to both you and "Robert Richmond" on the encoding
techniques u used,programs u used,resizing u used,options u used,why exactly those etc. but i won't.........as i said i can find it out
my self,also...your comparison should serve rookies to make their own comparisons....i have really learned a lot just by playing with codecs,changing settings,making bmp's etc.,yes,watching digital video sometimes too.....hehe]

but anyhow,you do have my respect and capturing guide is coming on nicely(as i said we on capturing forum will make it THE capturing guide).......and i still don't know of forum that would better suit my needs......

chill out ,man

/ivo

[int 21h]

*Usual int21h disclaimer applies here*

Color conversion processes introduce error and color shift into the original picture, this by definition would mean that the compression process of HuffYUV from a YV12 source is not 'lossless'.

These 'different' codecs are in reality the same algorithms with different parameters supplied, you could simply manipulate ONE of them and achieve the same result as manipulating 3 of them (with different names). (i.e. SBC encoding, which in fact does manipulate ONE .dll to produce the results of 2 .dlls of the past)

Keyframes usually have a higher compression ratio than a delta frame due to the fact that more information is stored in them, this being said, keyframes usually look worse than their delta frame counterparts. More keyframes = Worse quality, that was why we developed a sophisticated VKI algorithm from the base of Avery Lee's Vdub for Nandub (with the help of ViRA, thanks :))

Macroblocks are never incorrectly placed. The term macroblock is a measurement, a subdivision of a frame (usually 16x16 pixels). MV errors are something entirely seperate.

What test has this person made before Doom9 did? Doom9 has tested these codecs months and months before this other person did. And so did I, many, many months ago I had a Mpeg4v2 vs SBC shootout (unfortunately its no longer online).

**the int 21h disclaimer**: I've had a few Guinesses, a couple shots of Goldschlager, and some Smirnoff Apple Twist, its quite likely the text I type is no longer coherent. In the event that it isn't, I disavow any knowledge of typing it. In the event that it is, I take full credit.

[Doom9]

@^^-+I4004+-^^: Please make an effort to type more coherent and readable text. Every time I see one of your posts I immediately skip it because it looks like a tornado has gone over your text. What's with all those linebreaks all over the place? Type your text in the reply window of the forum and only press enter when you want to make a new paragraph. And if you type your text offline, use a program that automatically linewraps.. like wordpad (that's a program everybody has).

Now, I don't think I got excited. I didn't hand out a single strike, I merely asked for clarification and I think I deserve to know the whys when somebody comes in here critisizing my work. After all, I'm not doing it for myself but for you guys, and if there's something I could improve I'd like to know and that's only a natural feeling. You'd be feeling the same way.

[Selur]

Yup, I'd also like to know what makes doom9 guides "amateurish", okay e.g. the style of Nixed (http://koti.mbnet.fi/website/mpeg4test/) codec comparision is better, but I like doom9's comments about the codec more informing. And the comparison at the start is way of from beeing not "amateurish", in my personal view at least.

Cu Selur

Ps.: Would be cool to get a general answer what makes a codec comparison "amateurish" or not, so that one could learn a bit. ;)

[NeVeRLiFt]

Originally posted by PowerMacG4
@Neverlift

I hope you are joking...

http://216.239.37.100/search?q=cache:_wZrqbwVMPIC:www.adobe.com/motion/pcpreferred.html+&hl=en&start=1&ie=UTF-8

http://www.digitalvideoediting.com/2002/07_jul/features/cw_macvspc2.htm

[trbarry]

I don't have time to exhaustively test every last codec out there, any more than I'd have time to write every last piece of code I run. So I love these comparisons and always look forward to Doom9's (not at all amateurish) new efforts.

I don't have to believe every last opinion stated in any of them.

But I can still take away useful information that saves me a lot of time. For instance I mostly am re-encoding HDTV captures at higher bit rates. This makes my own desires maybe different from someone else's and my codec requirements different.

So based upon what I'm doing I can read a comparison like this one and casually say something like:

1) I won't bother to eval ON2 yet.

2) I'm glad I'm still using Xvid.

3) I wonder how that H264 thingy would do?

This is useful info to me.

- Tom

[ppera2]

Originally posted by Doom9
@^^-+I4004+-^^: Please make an effort to type more coherent and readable text. Every time I see one of your posts I immediately skip it because it looks like a tornado has gone over your text.

Well said :D

[Ookami]

@int 21h

I wish you would post more often and in more forums of this board!

Thanks, you are one of those people who really know what they're talking about.

BTW, remember our discussion on the old ezboard about the KF issue with vanilla Divx ;-) ?

[Doom9]

@selur: I think it's time I have something like that written for my next comparison. Preparations have actually started now, my computer is full of codecs I will probably never use again, but I don't have a timeframe for completion yet. I think the site you mentioned has a pretty nice approach, so if somebody could hack something similar for me, and include the possibility to have some comments for each frame (like you select a frame, and to the right there's a box where I could put the comment for the specific codec) and below the box I'd type the "after shot" comments like I'm doing now.

@tom: it looks to me like the vss h.264 beta2 codec might actually be fast enough to be included in the comparison. But hdot264 is definitely out of the picture.. it isn't usable except for real hardcore people with a lot of cpu time to spare.

[int 21h]

Originally posted by Ookami
@int 21h
BTW, remember our discussion on the old ezboard about the KF issue with vanilla Divx ;-) ?

That was truly 'back in the day' my friend ;)

[wmansir]

@doom9,

Since you mentioned you are starting preparations for next comparison I wanted to make a small suggestion. Do you think you could use 'mouse over' events for comparing your results? I've seen them used to great effect in recent video card reviews (like this one (http://www.anandtech.com/video/showdoc.html?i=1779&p=8)). It really helps people spot the differences in a sea of screen-caps that begin to look all the same, like the review that started this thread.

It would also keep it from looking so damn amateurish ;)

EDIT (before I even posted): I use a flash blocker, so I didn't see that Selur's link was a similar concept. Still, I think the use of mouse over events are easier to code, but could not provide the comment box function without incorporating it into the image or increasing code complexity. I also think the "original" from the DVD or Huffuv should be included in the set of images for comparison.

[Mole]

Just a small correction. I believe the threshold bitrate for auto keyframe insertion on scene change on the plain old DivX 3.11 was 400 and not 460.