as we started to discuss here http://www.hydrogenaudio.org/show.php/showtopic/16970 i believe doom9 was too quick to put ffvfw on the bottom of mpeg4 codec's list...

http://kotisivu.mtv3.fi/p.erse/

you have clips,you can inspect them till you go mad..i didn't saw xvid outperform the ffvfw as doom9 claims...but i have seen xvid's ME problems once again...

overall differences are very small in this clip,but they exist....

url simplified to the request of esteemed mr.Latexxx
(yes,Latexx check your dictionary now..hihihi)

sorry to say that, but a noisy, low quality capture clip, like you used, is not really suitable for a comparison imho

someone might come to the conclusion that a codec that leaves out details, produces a better image quality...

how come?in this test too one can see what's sharper and what's not......yes,this is 768x576 tv-cap converted to 512x384....and i have INTENTIONALLY left it unfiltered (normally i would use peachsmoother) to see what codec will handle SUCH video best....(actually i intended to produce artificial video:noise added to a static background,but why when i have tons of noises in my capturings..hehe)

however,i agree that there wil be people who will pick something that others will call too-smoothed...this comparison is for people who appreciate sharpness (as all of my previous tests)..i can't make chioces for people..i use mpeg4 flavours as they beat rv9 and vp6 with great margins if sharpness is concerned...

people who like blur can use rv9 as well!

do you have any real objections to the way i tested or is there something bothering you with results?
i saw you did some ffvfw too(on d9 forum..as you asked about could it be played on standalones),so what do YOU say?

on my tv-caps nandub and ffvfw stand a bit above the xvid..and if they do tv-caps OK,they'll do dvd-rips even better!
you know that's truth!

and if you have matrix dvd you should repeat doom9's test but with differentt ffvfw settings and different 2pass engine(ie. xvid's..beats me why doom9 used another one if no-one is using it..here it is straight from the horses mouth:
[quoting doom9]
"Now you'll argue that the codec has never crashed for you, but then again you'll be using XviD's 2 pass encoding mechanism, not the ffmpeg one (and we know XviD's 2 pass mechanism is good so there's no reason to test it twice)." )!

that's what i say....also,if you have better sources to test,send them over to me..i'll use anything and i believe ffvfw will beat xvid with any source...
and if you wan't quality,i'll capture something with no noise and will leave it at 768x576....would that satisfy?but xvid would fail even worse for such stuff....so i wonder if you really want that....

now,bond..tell me how is ffvfw worse than xvid?

first of all i have no interest in harming ffvfw or so (i love every opensource codec )...

yes, i did some small tests with ffvfw, cause i wanted to use its error resilience tools for xcd, but my (small) tests showed that it doesnt preserve sharpness as xvid does, it was only a small test, but enough to drop the idea of using it

hm, i understand that you wanted to use a sample that is useable to show which codec preserves more sharpness, but imho the capture is simply noise, i dont think its really visible to see what detail level each codec would preserve in normal circumstances, sorry

video is not sharpness only,you know....i too loved xvid's great sharpness performance,but it had visible ME problems that stopped me from using it....
if dvd-sources+xvid encodings don't exhibit those,then you're a lucky man...and i'm lucky because i have other codecs that handle noise better than xvid (as they have beter ME..)---i'm a tv capturer...i don't do dvd's......




this proves to me that you never captured a single minute of video.....so what do you know about it.....?
i only wish i captured matrix(it was rerun today here) so i could show you what does 768x576 tv-rip(encoded to ffvfw or nandub,no resizing) look like.....it'll beat any silly downsized dvd-rip,that's for sure!

but i'm not a "matrix" fan....i'm a "thin red line" fan....(although i don't get anime same as doom9... )

remember..the codec that does *noisy* video well can do anything!
no codecs handles *pure noise* well..(if you had capture card you would know that by now... )

i4004, those ME problems you claim to have spotted - you know that nobody ever managed to reproduce them.

In the spirit of the the 'solid-facts-only' attitude of this forum, I ask you to encode a clip and provide us with screenshots and detailed info on your setup so that we can reproduce what you are talking about. Otherwise please stop spreading that rumour of an 'XviD-ME problem' on this forum, too.

Your statements on the quality of diverse codecs seem very poorly rooted in facts and thus I personally think it is a shame if you doubt Doom9's painstakingly accurate way of conducting his tests.

Edited: my English grammar

i4004, I would very much like to see some really good examples of this ME problems in XviD. The motion estimation of XviD has been stable for quite some time IMO. I have heard you say this from time to time but I have never seen any really concrete examples. This qpel test is interesting don't get me wrong thank you very much for doing it, but I don't see any ME problem in the XviD clip. Also, I can't tell them apart in a "blind test." I there are a couple of things that exhibit bad performance for ME in all mpeg4 codecs that I have seen. The first is when there is a lot of dust/smoke. MPEG4 codecs completely bork on this kind of material and just turn it into a blockfest. The other example of bad performance is in low light scenes where there is a lot of noise. To be honest I am not totally sure whether this is a ME problem or whether it is a rate controller problem or whether it is a just the codec making a bad vop decision. At any rate almost all the noise gets removed and consequently a lot of detail. I tend to believe though that it is not a ME problem though and certainly ffvfw does a little bit better on this kind of material than XviD does. I dunno, though, noise is tough like you say commercial codecs like DivX5 try to cheat and remove a lot of noise and then add it back latter when decoding. Another thing I noticed about your ffvfw clip is that the first frame is bad. ffvfw builds seem to doing this lately, it may even affect compatibility I fear. If you are going to be doing a lot of encoding to ffvfw, the standalone ffmpeg might be a better choice since it doesn't have this issue. If you like a can provide you with a build, just PM me

It would help if you define 'ME-problem', i4004.

It is normal that noise is falsely detected as motion.

The only thing one could do about it is lower the ME's sensitivity (which is what cartoon mode does). But this is no good solution for non-cartoon material.

Usually (and sensibly) one applies a noise-filter.

[i have mixed quotes by tg and bonzi,but it should be easy enough to see which is which...tg never used ffvfw and bonzi did... ]


you didn's saw this
"don't inform me how ffvfw does a max. quantizer on 1st frame..i know about it,i don't mind that...."
on a web page?
http://kotisivu.mtv3.fi/p.erse/
it is under "other comments" section...



not true!i have heard neo-neko talk about it(actually we discussed it at vdub forum) and also,i have posted a link on doom9 forum where didee decsribed the VERY PROBLEM i hate about xvid!
so,am i alone?am i crazy?well,tg..am i?
( http://www.hydrogenaudio.org/forums/index....ndpost&p=163113 )
i have also heard few other people complaining about it (on my croatian NG...)
why do i have to repeat everything to you?
we had "nice" conversations back in doom9 forum days where i have clearly proved how xvid's ME is not top-notch....

if neo-neko saw it,didee saw it,i saw it,who are you to discredit our findings?
(perhaps it's only indicative that all of us (ie.neko,didee,and i) do tv-capturing...perhaps not....i know how xvid acts on my tv-rips ,and i'm not pleased with it..)

i have showed plenty of evidence tg.....nerly every encode i do has it....zoom into (zoomplayer or so...zplayer will remember zoom settings,so you can inspect all the clips with same area zoomed..) the area over the hill (as i described in the page) and you should be able to see it....also,there's that old encoding where this is visible as hell...
i will stop to say this as soon as i see it dissaper completely..ffvfw doesn't have it,nandub doesn't have it....xvid doesn't need to have it...

i also see xvid making rapid progress and i have mentioned that in my tests too...be aware of that also.......


blah,blah,blah!
i didn't said i doubt his methods,but it seems to me that doom9 got results that vary much from my results..(but true enough;
-i don't rip dvd's
-i don't ask for settings from codec dev's
-i don't watch stuff with deblock and dering filters....(so,did he used ffdshow deblock if he used xvid's???)
-i don't pack longer movies to 1cdr....)
i still say that he should do ffvfw justice;repeat the test,but with xvid 2pass engine,default settings on ME,and inform us on matrix used and if ffdshow was used to deblock and dering.....

your first sentence seems especially flawed;"diverse codecs"?well untill this i fully agreed with doom9 tests..(just look at my tests..they are almost identical to his,with the exception of ffvfw,and also he saw xvid in better light than i did in his last test..i fully agree with him on other codecs performance...reasons for this?;i believe he's biased towards xvid...and he really had the NEED to dump nandub by now--->now tell him that nandub's default DRF2-16 is CRAP!)..but it seems to me like this was the first time doom9 is using ffvfw....i'm trying ffvfw since it came out...
i told you before,don't take this as a personal inslut if i said i don't like xvid!


yes,i agree..neither can i!
so then,i have done my part!
you can't tell which is which and therefore you can't say which is xvid,ergo,you can't say which is better.....!!!
so you agree with me that ffvfw is NOT lousier than xvid (although in addition i say ffvfw is better... )!


well,if you think mpeg4 is bad,try mpeg2!
i completely disagree on this one;you should investigate what "quantizer" settings do for a codec!my dust/smoke looks OK...low-light scenes look OK here too...


this is xvid user speaking all right ( )


yes,keep the good word about ffvfw..that's truth and you know it!
i believe we're talking about the same xvid flaw..the one which is not in ffvfw...i saw this on fade-outs..which is essentially the same thing....

divx5?they are not serious folks!xvid is a jet compared to divx5...ffvfw is a jet with an afterburner,and nandub is a space-ship ( )


i'm not doing a lot of ffvfw (as i already said...still no-one can beat speed/quality ratio that nandub is giving...but ffvfw came closest..for my stuff!) but i'll PM you and i'll try that too!sure,i'll do anything!(can you provide xvid build that lacks the problems i'm experiencing? please give me that too! )

if there's an interest,i will capture some auto or moto racing (the things where i have noticed the poor ME job first) and will post those.....(but let me tell you,it WON'T look good on xvid!)
i need to hear two voices to do this;
tg should say yes,and bond should say yes........so there,if your xvid codec is bullet-proof they will say yes,i'll do the test and all will be settled!

either i'm the crazy one,or tg is talking nonsense...for example;



i would understand this,if other two codecs tested actted teh same way as xvid did!but they didn't!

anyhow,let me know about that auto/moto racing clips;i'll do my best to show you what i mean by "ME problems".....(once before "muf" said how indeed this was ME issue...
http://virtualdub.everwicked.com/index.php...&t=3614&hl=xvid
also,while you're there check out what are quantizers...
http://virtualdub.everwicked.com/index.php...44&hl=quantizer )

/ivo (willing and helpfull to xvid team... )

ps. tg,i believe you need higher sensitivity ME,or better decisions what bandwidth goes to DCT's and what one goes to MV's....(no expert me,but it seems like that..seems like xvid is saving too much on MV's)

That's quite a rant. Why do I have the feeling that someone's warning-meter will go up steadlily? Surely you haven't been banned from doom9.org for no reason.

And please do provide me with a link to neoneko's post where he describes an XviD ME-bug...

yes i have my friend!yes i have!(for no good reason!but i don't like non-free forums anyhow,so i don't belong there..that's the truth!i belong to US style forums..is this usa forum then?... )

neo-neko talked to me about xvid on a number of occasions...the threads were enormous and we talked about many things....

you wanna inspect these threads:
http://virtualdub.everwicked.com/index.php...d,i4004,and,bug

http://virtualdub.everwicked.com/index.php...d,i4004,and,bug

http://virtualdub.everwicked.com/index.php...d,i4004,and,bug
(lol,see the topic of that one)

and to be rather blunt,here's an excerpt from "making videos brighter" (quoting neo neko here)


(from http://virtualdub.everwicked.com/index.php...and,water&st=15 )
but wait a minute;why is neo's word better than didee's?you have not commented on didee's impressions on xvid's nastier sides!

also,i have called this a bug and i will continue to call it like that..it's a visual bug(bugs are the things that bother people,right?)
neko called it "swimmy effect",and that's a good choice of words too!
i believe we talked about same sort of artefacts....(it seems he saw it on encoding manga,as he does toons..)
it was acceptable to him..not for me....
later,he said he liked ffvfw (you'll probably find that too...)

again i'll find that too;

( http://virtualdub.everwicked.com/index.php...ffvfw,and,i4004 ) also,notice that i was not blown by ffvfw b-frames performance!

like this:

i still say that b-frames don't look so good in both xvid and ffvfw....(if i was to post that test sequence with b-frames you would agree...)
but laced encoding surely got MUCH better on both!

there..something to read for you guys....


but you skipped something (take up the gauntlet!);should i test xvid on moto-racing?
yes or no?

Ok first of all, lets not make this discussion personal. I am happy to say that for the most part it hasn't yet but I see that it could have the potential to get ugly. There isn't anything wrong with disagreement and even heated discussion as long as we respect the rules everyone will be happy and nobody will get banned. I think thats all that needs to be said about that.



Wups, sorry you are right



Nobody is trying to discredit anybody else we just want some nice concrete samples. Believe me, if you do this and there is something wrong it will be fixed in a jiffy. The XviD team is extremely receptive to finding ways to make the codec better.



Yes, I can't say which is better so yes I suppose you have accomplished what you wanted to.



Yup, if you search the doom9 forums you will find some posts by him about the same thing. You will also notice that I agreed wholeheartedly later in the thread. The trouble is that we simply do not know if it is a ME bug or anything. It could be something unrelated so lets not speculate. This swimmy effect is much worse with mpeg quantization which is why I only use h.263 quantization.



Ok, I can encode at constant quant 2 and it will still look very bad IMO, quant 1 is better of course.



Absolutely, if there is something wrong we want to know about it.

i4004, your reasoning needs serious review.

[edit:] Concerning Didée agreeing with you (in that thread): He does not. What he his talking about is known as 'moving walls effect'. Its probable reasons and possible cures were summed up by sysKin in an answer to Didée in that same thread:
This phenomenon seems different from your perceived problem. The look that you don't like about XviD is the effect of extreme noise in a static are which cannot be distinguished from motion and also quantizes very badly. The only sane thing is to use spatial/temporal noise-filters because they can and should distinguish between noise and motion. An MPEG4-codec does not have the task of going spatial or temporal in order to recognize noise, nice as it would be. And MPEG4 by its layout does not do prefiltering or in-loop filtering, either. And never will. You should use RV9 if you want a codec with integrated prefiltering. What Didée and sysKin talk about is a flat-coloured (preferably dark, as Didée describes it), still area which seems to move around the contours of a moving foreground-object. In your sample the background is anything but still; it's sea of noise. And what you take for 'tearing motion' looks much more like a typical ringing effect which quarterpel tends to amplify even more. If you had properly noise-filtered your source and that 'tearing motion' would have appeared, you could attribute it to that 'moving walls'-effect. Try it and we shall talk again.

I respect that you don't like how XviD treats noise but I cannot let it pass if you wrongly call it a 'bug' of its motion estimation. Simply because it is not. Full stop. My last edit (hopefully) to this post: the 'moving walls effect' has not been called a bug, it is more like the infamous 'blocking in dark areas' (which seems to have been solved by now): something to be improved, not something broken. Though the difference is just a gradual one: whether something goes wrong absolutely (inverted-luminance macroblocks should never happen) or just has a tendency to look wrong (sometimes XviD has difficulties distinguishing luminance fluctuations from motion).[/edit]


More importantly, you use references to strengthen your point that actually weaken it. In the threads you link to, Neo Neko did not agree with you.

quoting Neo Neko from http://virtualdub.everwicked.com/index.php...d,i4004,and,bug:

again (ibid.):and again (ibid.): and yet again (ibid.):
finally Neo Neko attaining another reason for what you call a 'bug' (ibid.):
I whole-heartedly agree with Neo Neko here.

Now,in the second thread you point to (http://virtualdub.everwicked.com/index.php...d,i4004,and,bug) Neo Neko does not talk about ME at all but rather about not having the time to talk about it. And the same goes for the continuation of that thread.

The third thread (http://virtualdub.everwicked.com/index.php...d,i4004,and,bug):
Here is a very nice quote from Neo Neko, answering another repetition of your claim that you actually found an ME-bug:


What we see from these threads is that not only does Neo Neko not agree with you - but as most senior users on doom9.org he also has a problem with your way of reasoning. That is because most of us have learned the rules of scientific reasoning at some point during our education. This is what makes for the high level of discussions on forum.doom9.org and enforcing a fact-based argumentation is what keeps doom9.org a place for these users.

Your reasoning on the other hand is really make-belief. And we won't have that. Not only is it irritating for other users, but it dilutes the fact-finding process which leads to actual improvements in open-source software. Your way of incoherent argumentation also and inevitably triggers reactions from users like myself which leads discussions to drift off-topic - for which I apologize.

Now, a forum is not a university. If you delivered a paper in that style for review to your professor, he would not only criticize that your quoted sources don't prove what they claim to, but he would also become suspicious that you hoped the reader would be too lazy to check your sources. Your attempt to strengthen your disputed claim by quoting a well-respected authority would be called 'misleading' because that authority hasn't agreed with you in a single line, but quite strongly disagreed with you.

Well, as I said above, a forum certainly is not a university, and that's a good thing, too. But on the other hand it makes violations of a scientific argumentation style worse because people don't expect other people trying to mislead them on a forum. So usually statements like yours just go unchecked and thus false rumours disguised as 'fact' are created. It's just because I already know you that I checked your 'proofs' (clips) and sources.

Last not least it is a disgrace that you should misuse the reputation of people like Neo Neko. He certainly would not like this. And you know the place where I would find myself forced to react to this by using rather drastic measures.

Yours,
Tee

{gentleman,start your fire-extinguishers! }

[quote]Ok first of all, lets not make this discussion personal[/quote]
i'm doing my best!but i'm struggling!
it's easy to talk to people like you,bonzi,but tg got some very wrong impression about me,but i'll try to stay calm...(although i won't be surprised if i loose it..)

[quote]The XviD team is extremely receptive to finding ways to make the codec better.[/quote]
nope,it isn't.....nic said(way back) that "they will fix it if they find it to be a problem" (ie. he says xvid is essentially ok,and i'm essentially a nut... )

[quote]Yes, I can't say which is better so yes I suppose you have accomplished what you wanted to.[/quote]
thanks!i appreciate it!(also,you can't tell nandub apart from others,can't you? )

[quote]You will also notice that I agreed whole heartedly later in the thread.[/quote]
i never perceived you as someone who disagrees with me;for crying out loud you used ffvfw man!that speaks volumes to me!we need more people TO USE FFVFW!
and compare it to XVID!

[quote]Ok, I can encode at constant quant 2 and it will still look very bad IMO, quant 1 is better of course.[/quote]
my troubles with mpeg's are usual;fast motion needs to much of a bitrate,and gradients tend to block(but THESE ARE NOT MISTAKES I'M ADDRESSING!)
i'm doing fine on water,fog,fire...etc.it looks fine!off-course it won't look fine if fire frame was compressed with quantizer 16....sure it won't(that goes for water and for too)!
but blue sky needs real-low quantizers and i can't afford those....and i can't think of a way to fix this 8x8 DCT block problem for gradients....

[quote]The trouble is that we simply do not know if it is a ME bug or anything.[/quote]
i'm using word "bug" in a wider sense!it's not like it's crashing xvid,but it doesn't look as good as ffvfw or nandub!ie. it's something like divx3 "shits",but not so pronounced...(shit manifests in many way,some of which look VERY ugly...xvid's problems are less than "shit",but very annoying too!)

[quote]Absolutely, if there is something wrong we want to know about it. [/quote]

i have asked tg and bond (declared "defenders" of xvid..) if they both say yes,i'm doing it....(although YOU will receive such .avi in no time..it's like 1,2,3 for me to make xvid flop on motion!i will probably make it today...if i have time enough)
i and tg had disputes long time ago....we need to settle it in this way!
he will ask from me and i will show him how hi-motitonis destroyed by xvid!
as i have my pride too!


tg;

[quote]Concerning Didée agreeing with you (in that thread): He does not. What he his talking about is known as 'moving walls effect'. Its probable reasons and possible cures were summed up by sysKin in an answer to Didée in that same thread: [/quote]
have you read it?he talks about 3 DIFFERENT FLAVOURS OF FLOATING WALL!
one of those flavours is the thing i described!
here:
[quote]3.
"Tearing motion":
Imagine you have a pretty dark scene. Something in the foreground is in focus, the background is almost flat colored (black sky, e.g.).
Now, if the thing in the foreground starts moving, the background starts moving also (!), as if it was "teared" by the foreground object.
This should be resolveable by a more sophisticated creation of motion vectors (don't create and trust predictors too blindly). However, that's faaar beyond my horizon.[/quote]

notice that he mentions MV's too(same as i) but also,this is beyond my haed too...i'm not codec dev to tell you exaclty what you need to change!
but why don't u use ffmpeg's motion estimation engine?blow to xvid team's pride?well......i would give up if i can't make something work in what...2-3 years?



[quote]The look that you don't like about XviD is the effect of extreme noise in a static are which cannot be distinguished from motion and also quantizes very badly.[/quote]
there' was no "extreme noise",but yes,generally you are talking business now!
xvid is doing poor ME decisions and smears and "swims" some portions of the frame!this is NOT limited to walls (gradients) as that tis a pain for any mpeg system,but "tearing" (i call it jelly-waving,neo calls it "swimmy effect") si a real problem!
in my mind,MV's are getting less bandwidth than they should!
i WOULDN'T MIND if i didn't saw this work OK in nadub(and divx3 is VERY outdated codec) and ffvfw (fabrice ballard MUST be true when he says they built it from scratch when i see it act diferent(better) than xvid),,to quote didee again
[quote](don't create and trust predictors too blindly). [/quote]

[quote]The only sane thing is to use spatial/temporal noise-filters because they can and should distinguish between noise and motion.[/quote]
why?i will rateher use a codec that handles it better than xvid....
spatial filtering blurs (i HATE blur),and temporal ghosts (i don't like ghosts either),so if a codec can swallow a noise better than xvid,it won...nadub won,ffvfw won....

[quote]And MPEG4 by its layout does not do prefiltering or in-loop filtering, either.[/quote]
YES!OFFCOURSE! this is why i appreciate mpeg4 so much!
i did wmv9 and it's obvious it has inloop stuff...so i'm eager to see h.264..it has been said it has it too,but i hope it can be turned off manually!
it is also obvious (i dunno how many codecs you tested) that old guard of h.261 has it (indeo derivatives etc.)...when i read specs for h.261 i see it has in-loop filtering...when i read h.263 specs i see it doesn't...
i HATE in-loop!
one can achieve BETTER image quality with sharp image,no in-loop,some blocks on playback but with additin of noise on playback (i compared divx3 at 320kbit to wmv9 on 320kbit.......with ffdshow noise additon i have made divx3 look BETTER...i have added same amount of noise to wmv,but it was still too blurry....!)
codec dev's should look for bettetr ways to achieve efficiency,and NOT by silly in-loop tricks....this also proves one more thing;mpeg reached it's limits,so filtering is applied to improve the efficiency...i DON'T LIKE THAT APPROACH!
we DON'T WANNA HAVE RV LIKE MPEG!!!



[quote]What Didée and sysKin talk about is a flat-coloured (preferably dark, as Didée describes it), still area which seems to move around the contours of a moving foreground-object. [/quote]
didee described 3 DIFFERENT MANIFESTATIONS of same issue;i call it poor ME performance.....
(and i see now you saw it too...as you decribed it well few lines up)
you have described it well here!
that's my objection to xvid!

also didee said this too;
BTW, while we're at artefacts:
[quote]The floating is more pronounced with ffdshow than with xvid.ax.[/quote]
and that's exactly what i saw too!xvid decoder did a better job than ffdshow!
to follow up on his thoughts,does this mean that xvid has non standard mpeg4 decoder,or ffdshow has it?


[quote]If you had properly noise-filtered your source and that 'tearing motion' would have appeared, you could attribute it to that 'moving walls'-effect. Try it and we shall talk again.[/quote]
i most certainly WON'T!
you fail to grasp the most important thing;read my lips:
FFVFW AND NANDUB DON'T EXHIBIT SUCH POOR BEHAVIOUR ON MOTION!!!
OK?
the filtering won't help to xvid,as then ghosting will put it in REAL BAD SHAPE on motion (as ghosting is much more demanding than noise alone!)ghosting is FINER NOISE THAN NORMAL NOISE!!!
have you ever used temporal smoothing?temporalsmoother() from dividee's mpeg2dec.dll,or peachsmoother or likes?
if you had you would know btter (c3d or fluxsmooth(if that's the name..i forgot the unefective filters) don't realy do anything for analog noise,and i'm ALWAYS surprised to see that u folks use it for dvd rips....this means that mpeg4 codecs are not as effective as you would like,OR you're pushing them too hard so you need extra smoothing...(i can't believe that doom9 puts matrix(over2h) on 1 cdr....i think that looks like lo-res crap..honestly...) mpeg4 is better than mpeg2,but IT'S NOT 2X BETTER!!! fabrice ballard speaks about 10-20% (to quote him:
[quote]Note that MPEG1, MPEG2 and mpeg4 are very close. MPEG4 gives only
marginal improvements to MPEG1 (10 or 20%).[/quote]

i say he's right!if you do mpeg4 with 30% lower bitrate,then it'll look worse than mpeg1/2....if you do 10-20% less bitrate,then it'll probably look simillar!
test is test when it's done on same resolution/same target bitrate for both....you CANNOT compare downsized dvd-rip to 720x576 original...(and i see doom9 uses biliner(crap) and that he used worst deinterlacer available(fielddeinterlace).....i have helped develop "kerneldeinterlace" (i have suggested to graft what components should incorporate..would you like a link for that too?) and it's the BEST AND FASTEST deinterlacer....EVER....so why is doom9 so uneducated about such issues.......bicubic resizing is a must,as bilinear smooths too much...)

beats me....

[quote]
I respect that you don't like how XviD treats noise but I cannot let it pass if you wrongly call it a 'bug' of its motion estimation. Simply because it is not. Full stop.[/quote]
i REALLY don't care how should we call it!
i said it's an annoying visual-bug that's not present in ffvfw or nandub(divx3)!
i feel that way!
ME should be done in a better way if you want to be a No.1 codec.(Full Stop)
ME CAN BE DONE BETTER!ffvfw and divx3 PROVED IT(divx3 long time ago,and ffvfw just recently...i dunno about 3ivx,but i hope they did a better job than xvid team....)
when this annoyance is no more,when asked about mpeg4 i'll prefer xvid...(but,if xvid gives higher bandwidth to MV's will it be as sharp?you cannot have both,you know...)

[quote]something to be improved, not something broken.[/quote]
well,yes!
sure!
i didn't said xvid is a broken thing,but just that i don't like some things about it!
i'm always saying that xvid is much better mpeg4 than divx5.....(divx5 has simillar ME issues,but it's also too blurry....again,remember..ffvfw and nandub do a smashing job on ME!real good!)


[quote]Though the difference is just a gradual one: whether something goes wrong absolutely (inverted-luminance macroblocks should never happen) or just has a tendency to look wrong (sometimes XviD has difficulties distinguishing luminance fluctuations from motion).[/[/quote]

YES!you're right and even in this post i have said that "shit" can look worse than xvid's issues!!!
inverted-luma blocks are..well..it can't get much worse than that...BUT i don't have them with nandub!if i saw them,i would convert to xvid immediately.....(there was a discussion why they appear..some said poor RAM,some said app's,some said this,and some that..i don't care:i don't have them;i have intel system with siemens/twinmos ram on it...no luma inverting here!..nandub never made such a mistake for me!and i even don't turn on anti-shit...)
[in the end we'll agree...huh.. ]

[quote]More importantly, you use references to strengthen your point that actually weaken it. In the threads you link to, Neo Neko did not agree with you. [/quote]

me and neo-neko generally don't agree on many things,but he said he saw "swimmy effect" and he said ffvfw was surprisingly good.....
to a clever man,this is enough....

we CAN discuss if i should call this a bug,or a "different way of doing ME",sure we can,but i call it like i call it,because it is VERY annoying;i watch stuff on tv-out and i can clearly see it even there!
FOR ME ,BUG,FOR YOU "DIFFERENT WAY OF ME".......but at least now you know what i mean by "bug" and no further misunderstandings are possible.....my "bug" is your "different way of ME".....clear enough........

[quote]QUOTE 
The swimmyness was rather uniform but the image over all did retain it's shape nicely. Anyhow chalk it up to early MPEG quant implementation.


I whole-heartedly agree with Neo Neko here.[/quote]

i do too(about "early implementation) i don't agree that "shape is retained"...it's not,it's messed,background is swimming heavily,and it's tearing(if this is his comment about xvidME clip....)!
BUT why is xvid using "early MPEG quant".....why i see same things as i saw in divx4?why haven't you rewrote momusys completely????
WHY?

[quote]Neo Neko does not talk about ME at all but rather about not having the time to talk about it. And the same goes for the continuation of that thread.[/quote]
yes,i haven't checked all of it,those are long threads!
but it's true,we didn't continued it...how could we when he saw problems too....he called it "swimmy effect",remember?
and he's not codec dev so he can tell me more about it....

[quote]This is what makes for the high level of discussions on forum.doom9.org and enforcing a fact-based argumentation is what keeps doom9.org a place for these users. [/quote]
if your "school"(btw. that's VERY FUNNY argument!) tell you that xvid looks good (inspite what your eyes tell you) then,again,you're a lucky man!
you have find a nice quote by neko there and i fully agree with his description!!!
RIGHT!I(me ,myself and i) THINK IT LOOKS NASTY!

but tell me about didee too!
i respect him more than neko.....why did he found 3 objections(3 manifestations of same issue) if xvid is so smashing?

[quote]Your way of incoherent argumentation[/quote]
no,i just post clips and samples of what i dislike!
if what you saw is OK with you,then discard my stuff!!!

i WON'T ASK OF YOU TO TELL ME IF SOMETHING IS LOOKING GOOD OR NOT!be aware of it!i have eyes too,you know....

[quote]And we won't have that.[/quote]
who's "we"?are you a mod here too?
let me telly ya,buddy;i'm not afraid of you and i'm not afraid of your threats;i will be only glad to be thrown out from the forum that is threatening because i don't like some of the xvid's stuff!!!
ONLY GLAD!!!
keep you mod threats for doom9..they are well thought of there!
keep you users silent..real silent;that way xvid can stay like before!

is this rant then?no,you're provoking me!nobody threatens me!NO ONE!
ars,neuron2,vdub,hydrogen,doom9...no one!
MAN I'M TRYING TO HELP YOU!
don't tell me that's crazy thing to do!
i WAN'T TO SEE XVID GO BETTER;GO FASTER;GO SHARPER!
nothing else!

[quote]Last not least it is a disgrace that you should misuse the reputation of people like Neo Neko. [/quote]
disgrace?how?because i quoted him?
i have MUCH deeper respect towards him then towards you;you are actually using your d9 forum mod status to talk silly again:he doesn't..we disagree on MANY things,but i respect his opinions...he tries to post some arguments too,but all you talk is..what?scholar apprach?university?what?

if he doesn't like this,he will come here and tell it ,or he will tell me and i will quote his words!like a man to man conversation,and not trying to say i'm nuts because i hate some xvid's ways....

anyhow i'm glad to see that you also saw xvid's issues on some things,so my goal is completed!
issues exist,and xvid team has some more work ahead of it!


you haven't answered about moto racing still!
i will send test clip to bonzy and i'll see what he says(i won't listen a person that used xvid only and still manages to try to defend xvid..against something he NEVER tried!!!LOL!)
i will make nandub,xvid and divx4 to be rather blunt about xvid roots!
(but i also have a salma hayek movie today on my menu,and that MUST( ) be captured too...)
good codec handles EVERYTHING WELL!remember that!

cheerio!

/ivo

[mods are invited to correct my post if they see the need,but let me tell ya,i won't like that....and then i will require that tg's post would be modified too against hints and allegations he said against me!]

PEACE BROTHERS!

I will not warn you guys again any more violations of the rules and there will be consequences. I would like everyone to read this Rule # 2 again before posting. Next violation of this rule in this thread and I will lock it and there may be other consequences to follow.



Actually, I can. As much as I fear saying it, it does look better than the XviD and ffvfw with qpel. But if you compare it against the ffvfw and XviD w/o qpel, I can't tell the difference.



Well, no first of all nic is not really that involved in XviD development ATM and there are many others in the team as well. And I don't think it is fair to assume that he meant you were a nut case from such a polite statement.



Well that isn't fair either. First of all, lets not insult all the hard work put into XviD and lets not forget that for many people XviD works very well. Chances are many people don't use it to encode such noisy material. That said, I really can't tell any difference between the XviD and ffvfw w/o qpel and nandub. Possibly these issues are more related to qpel than anything. I can't tell you how many times I have run across noisy samples which completely kill all qpel implementations. As soon as it is turned off though, it is much much better.



It does have inloop filtering but at least in the reference encoder this can be turned off. Obviously it will depend on the implementation but certainly if there were an open source implementation most likely there would be an option to turn it off. I personally hate inloop filtering too, I would much rather have some blocks and other artifacts than have a blurred smoothed unnatural picture.



Ok, first of all these statements are from a while ago before the 1.0 betas which I think everyone will agree are much much better in this area, sure it isn't perfect but it is considerably better.



hmm, but I think this is what Teegedeck is trying to get across the codecs mistake the noise for motion and this leads to bad artifacting. But filtering the noise will lead to other unpleasant artifacts though you are right so that certainly is not a solution. That said, I know Teegedeck is not at all a fan of filtering just take a look over at doom9 and you will see what I mean.



Please these personal attacks are not acceptable. I would like to continue this discussion but I will not if this keeps up. Please tone it down a little and don't SHOUT



Ok, again. It has gotten personal and very rude. This is the kind of stuff nobody appreciates. I think if this discussion is to continue any further it must be scientifically as Teegedeck described. I think i4004, your qpel test was very scientific and I am sure your motoracing samples will be as well so lets continue this discussion in the same manner.

Please, calm down everyone! There really is no need for this thread to turn in a personal and possibly insulting direction. This is an interesting topic about how good ffvfw really is.


I have to ask you for evidence for this statement. At least for audio coding, it is a well known fact, that a codec's performance with test signals and music are two very different things. And I do believe, that the same holds true for video signals.

About the samples: While I am as well using these codecs for encoding TV-captures and am thus interested, how good not perfect signals are encoded, I think that this particular sample really is too noisy for a good comparison. It is difficult to spot a quality difference when both videos suffer from severe artifacts throughout every frame.

I would suggest that you use a denoiser for your next test (BTW, what Peachsmoother settings are you using?), so the codecs are tested under more "normal" (as you would use them "everyday") conditions.

I think there was a clear difference in your arte-screenshots (linked in another thread) between Xvid and ffvfw, I would like to see a clip more like that, i.e. that shows the same deficiency.

I didn't read all the topic, so maybe this was already mentioned here, but this topic is based on false assumption: If A codec is better on noisy material than B codec, it doesn't mean it's always better in all circumstances. Rududu codec is simply unbeatable on very difficult, noisy video, but it isn't really competitive in DVD-rips.

Xvid doesn't handle noise well, but it was optimized for clean sources. It shines there and the annoying floating background rarely occurs in that case. FFVFW and SBR are more error-tolerant, so the artifacts are less annoying, but on clean source, xvid is really better than them (and this is not assumption, but experience ), because it is specialized for encoding DVDs, while the other codecs are more for general use.

CALM DOWN
new movies will not have noise - xvid is better here??? I donno


I will help you guys to make a conclution
DVD rip -XVID
TV cap - ffvfw

heheheheheheheh, I never do Video encoding before,

well,yes i can live with such explanations also!
i only know that xvid doesn't like my caps...i never did dvd-rip....
so it seems to me some consensus was agreed upon:ffvfw and divx3 take noise better than xvid.......
(i believe we all agreed on that one(?))
and therefore we could call xvid an dvd-rippers codec...

so if it's ok with bonzi i might as well skip the moto test,as we reached an agreement....(i just did a quick moto test,and yes,xvid didn't liked it particularly....)
[and if i see problems with xvid dvd-rips(that other people did),i'll let you know..hehe]

tommy,any wavelet style codec acts as blur filter(vp6,rv9,rududu) so noisy stuff gets to be denoised...but this destroys the details also,so i wouldn't agree that this is what we need for noisy stuff.....(i may as well use spatial filtering+mpeg4 to achieve simillar results...huh)

continuum;

some stuff i filter,and some stuff i don't....(it's not like i gave xvid something harder than i give to nandub every day)...
i use peachsmoother in a way to first assess the noise ;

then when peach spittsout setting (it needs still frames tot asses the noise;bare in mind that!) ;


etc.(here peach was applied after the resizer...)
this is manual way of doing it,but you can also use automatic way if you wish
like this;

noise estimation is internal/automatic,but denoising tresholds are applied...



well we all seem to agree that xvid is worse on analog noisy stuff,so no wonder ffvfw beats it there...if you're doing capturing,you can produce many such clips...(i'm not putting you off,but usually once i find a better codec for a particular pirpose,i stick to it)

also big thanks goes to bonzi,and needless to say i agree with all his remarks in his last post!
RESPECT!
if anyone needs anything else,let me know....(i still didn't tried that standalone ff encoder i got )

/ivo

Real video 9 isn't wavelet based as stated by their codec designer many times at Doom9's forum. Rududu is indeed a wavelet codec. Vp6 doesn't look like wavelet codec because it has dct blocks.

notice that i have said "wavelet style"!
i know karl lillevold said rv9 is not wavelet......

"wavelet style" for me means "blurier than mpeg"....
(all three mentioned codecs have that in common;they blur)

i didn't said that rv9 and vp6 are wavelets,but they blur like wavelet..

mpeg style si sharp style(hehe)

i will also add a link to doom9 forum where a guy says that noisy "das boot" did not went well with xvid....fits this thread perfectly,i believe....
guy talks about "swimming artefacts" on a noisy film....

http://forum.doom9.org/showthread.php?threadid=67863

[offcourse,i disagree with koepi that other codecs wil exhibit simillar problems...we know the ones that do noise better than xvid.. ]

wavelet doesn't imply blurring...

if a source is flat with noise, there is very very little low frequencies, but more high frequencies. so theoretically one could gain an equal compression by quantizing the low end rather than the high end.

wavelet codecs are in a better position to do this because the transform is applied to the whole image, not 8x8 blocks. also, DCT codecs have VLC's applied to them which are most efficient when the coefficients are highest at the low end.

would you care to show some proof that wavelet doesn't blur?
send in 2 images;jpeg and jpeg2000...make them both with low compression ratio (ie. like jpeg@90% quality or so...)....

so far,i never saw sharp wavelet,but perhaps you did(?)

but i never toyed with wavelets that much anyhow.......if you can prove that wavelet can look sharper than dct,i'm listening....

Ok, guys I came across a interesting thread which has some relevance to this thread, I direct you to gruel's second post here http://www.xvid.org/modules.php?op=modload...iewtopic&t=1781 very interesting indeed .

uuhhmm?
they discussed GMC there with very lil mention of normal MC/ME...

although there was this:

by Celebrandil

also; i never used GMC in my tests....

No read further this is what gruel said,

oups,missed that;yes,you're right......

this stuff was *constantly* on my mind;i asked myself again and again:can it be that the codec that does noisy stuff well(divx3,ffvfw) will be worse(in encoding a clean source) than a codec that does noisy stuff in a lousy way(xvid).....

it didn't really clicked in my mind;expected behaviour would be to expect same or somewhat better performances from a codecs that also do noisy stuff better....

recently(10-15 days ago) i got dvb receiver,and therefore,a source of much cleaner video than analog sources.....
the signal is still capped via bt8x8 capture card(@768x576),but there's no noise to talk about anymore....

so i stacked xvid(RC4) against divx3 once more......
it is 1,8mbit/s,1pass,p-quant 2-7 comparison....
xvid has all the "advanced" options turned off(ie qpel,gmc,adaptive quant.,b-frames..etc.),highest motion search precision("6"),vhq off....
divx3 doesn't have any of these options( )..divx3 is used via nandub,xvid used via vdub 1.4.10.....

looking at both clips at 25fps,viewer will have enormous problems in differentiating the two..i think that even experienced encoders couldn't tell these two apart....so one thing must be said;nandub encodes faster:my previous tests with xvid RC4 showed that nandub is considerably faster even if i turn off vhq and bring motion search precision to "5"......

again,my logic was ok;it would be unexpectable to see a codec that does a poor job on noisy sources to do completely opposite(ie excellent) on clean source....(because divx3 will also profit from a clean source!)

now everyone asks,wtf didn't i used advanced options of xvid?
there's only one answer;speed.....
i'm looking for fastest and best encoder....not slowest......i encode a lot of stuff and i can't afford a slow codec...
also,i already did qpel test,and i didn't convert to qpel at all....gains in quality aren't that big so qpel would pay-off.....
(i tried to employ other options too,but none of them is a magic that will make xvid look awesome....or,to say it differently, visually better than divx3)

in other words,if nandub is doing 25fps encoding,and xvid is doing 20(or less) while image quality is such that i can't differentiate the two,i will use faster encoder....
that is the simple rule i follow;if quality is equal(or differences are so small that i can't differentiate the two) i will use faster encoder...as simple as that....

in this respect xvid is not a clear winner at all;it is
1-slower than nandub
2-it doesn't produce better looking images(when we match the settings of xvid to nandub's...match it because we wanna see if xvid is faster or not)

i need either same speed(as nandub) and better image quality,or same quality(as nandub) and faster encoding....(or,offcourse,both at the same time)


i will be more than happy if someone can provide an xvid settings that will result in one of these conditions fulfillment....(one will be quite enough)

DivXc32.dll [ 01.04.2000., 5:35:06 copyright microsoft corp.1996-1999.,cracked by gej]
xvidcore.dll[ 05.04.2004., 7:36:48 copyright ???]


grab some images here
http://kotisivu.mtv3.fi/ii00i/xvid-rc4/xvid%20rc4.rar (1,16mb)

to connect this to ffvfw too(as it's in the thread subject...hehe) see this (again)
http://i4004.bizhosting.com/ffvfw_the%20quest/

Newer versions of XviD compiled from the branch which will eventually become 1.1 perform somewhat better on noise now since sysKin added a better skip detection. So a lot of the unpleasant artifacts such as the moving walls are much less. Also, you might think that with skipping more macroblocks less noise will be preserved; IMO this appears to not be the case. The human eyes seem to react much more positively to sort of uniform random noise rather than unnatural noise such as the moving walls effect. Also on a unrelated note, I am curious as to how or why XviD is slower than nandub. But then again I never encoded with nandub much however I remember divx4 was noticably slower when it came out. XviD with fast first pass activated should be very fast. I can get about 30-35fps with fast first pass on p4 1.8ghz. By turning everything off like you did, you should get about the same as fast first pass.

the problem with noise is such;if you master an old movie onto dvd,it has noise....so it's not like xvid can now be called "dvd-rippers codec" to full extent......that is the biggest problem i see with xvid....it's not universal codec...and after all we cannot influence our sources that much....if we have dirty sources,they still need to be encoded...me and wilbert did our best to clean some noisy source enough so that xvid likes it,but we failed.....still we saw typical xvid/divx5 problems....

i indeed believe that xvid works totally acceptable on clean sources....
in this test,i didn't notice any xvid/divx5 noise problems...(as there's no noise)


i've been saying that ever since i saw my first xvid encoding...
that sort of handling noise is unnatural..i doubt this is an intentional flaw,as no one would intentionally make such mess...i see it more of an problem that stayed with xvid longer than it should of....


i think this can be attributed to divx3 being of microsoft origin....
back when it was written,it needed to be fast as it was run on slow machines....now it's speed just exploded....while xvid was never that heavily optimized......nandub was always faster......regarding that issue,xvid team should also learn from ffvfw,as ffvfw can be setup in a way to match nandub's speed......
(ffvfw source also holds an answer on how to handle noise properly)

speed again;in that speed test i talked about,where i set the motion search precision to "5",i was already starting to see xvid's performance deteriorate in comparison to nandub's....decreasing it further would probably result in xvid being closer and closer to nandub if we look at speed,BUT by the time it went as fast as nandub,it would look much worse.....
(in my mind already on "5" it looked worse....nandub was cca. 5fps(!) faster against such xvid setup...i was encoding some lores video;nandub went 25 and xvid 20fps...)