Codec Comparison Test

[Satirik]

hum encoding anime movie like futurama with a bitrate of 987kbps for a width of 604 is stupid coz u cant see the problem with the codec (because there is less detail on anime than on movies) ,try to encode the same movie without resize (just crop) , with an averange bitrate of 400 or less and you will see than this comparison is not optimised although it's writen than the size and the high bitrate are not the best for this kind of test ...

because for anime RV9 is better than xvid with low bitrate ...

a screen i have made on a movie a month ago to compare xvid and RV9 , size 716 bitrate 200 :D

same size at the end ... and you can say what you want there is so much detail in the RV9 picture (on the bottom) than in the xvid one (on the top)

http://0.t.free.fr/rv9vsxvid.jpg

(sorry for my poor english ;))

[BlackSun]

omg not you :) You speak better in english than in french

[TelemachusMH]

Anime/Cartoon sources are actually harder to compress and they take more space than regular video. The reason for this is because of the sharp changes in color (aka lines).

I would agree with you that RV9 does much better at low bitrates, but even looking at your pic, neighter of them are that great. The codec comparison was comparing at average bitrates that most people use. If you want to make another comparison for low bitrate encodes, go right ahead. I'm sure many people would be interested in it (including me). But please don't just critisize the work Doom9 did. It doesn't help any, and is very discouraging. I for one am very glad that he included Futurama in the comparision. Lots of people use the codecs for anime/cartoon videos, and they haven't been in any of the previous comparisions. It was very interesting to me to see the results of it.

Thanks Doom9!

TelemachusMH

[Satirik]

never critisize the doom9 work :) i just posted because i think there is a problem and i wil probably make a complete comparison for "anime and codecs" . Until now every test doom9 did was the reference in the divx community (for me too for sure) but i think codec compression compress better and more so why do not use them to their maximum :)

i will make a test (only for anime) , i will select a chapter of a recent dvd anime and i will encode (with last codecs) in :

320 (maybe i won't do it but i think it can be very interresting because i lost a lot of detail but win a lot of place and noone watch tv to in front of the screen so you can't see the detail although they are)
512
576
640
720 (or a little bit less because fo the crop)

with these bitrate :

200
400
600
800
1000

so a total of 25 test :) but the problem is than i have no recent anime dvd ... but i will do it ...

[JoeShrubbery]

Originally posted by TelemachusMH
Anime/Cartoon sources are actually harder to compress and they take more space than regular video. The reason for this is because of the sharp changes in color (aka lines).


I suppose that might be true if you're working from a really good DVD transfer, but if you're working from an analog capture then in my experience that's totally off base. I've been capturing TV shows, both live action and animated, on a regular basis for the better part of the past two years. I know I'm not the most experienced in the field, but I'm hardly a newbie either. Whether you take the raw capture, or do some filtering before hand, on average I've found cartoons much more compressible than live action content.

Case in point, I've been capping Clone High off of Teletoon up here in Canada, and after some careful filtering (including Flaxen's cartoon filter for Vdub to sharpen and enhance edge lines) to compressing with xvid constant quant 2 yields a final compressed video size of between ~240-280 megs @ 512x384, depending on the episode. If I try compressing constant quant 2 on an episode of Trailer Park Boys the filesize balloons to well over 400 megs. Granted, Clone High gets IVTC'd whereas I just deinterlace TPB (it's shot on DV, not film) but even so it's only running with an extra fifth the frames and yet if compressed using identical settings easily uses over a quarter (more like a third) more bits. Even if you compress without any filtering whatsoever, save deinterlacing or IVTCing, I always found animated content either demanding less bits than live action or coming out less blocky and/or smeared (depends on the codec) when using the same bitrate.

[Doom9]

Bits/pixel*frame values:
Matrix: 0.139198
SPR: 0.188076
Futurama: 0.140030

I calculated those values before chosing a resolution for Futurama, which is by the way a very clean animated source, not your average noisy shitty transfer NTSC anime edition (which will never make it in the test because I don't like anime). So while you might argue that Futurama got a too high bits/pixel*frame ratio, using the next higher resolution would've resulted in a 0.128594 ratio which I consider pretty darned low no matter what the source is.

@Satirik: XviD has not been optimized for low bitrate, whereas the streaming oriented RV9 has. So, it doesn't make much sense in comparing XviD and RV9 at 200kbit/s, it's clear that XviD can only lose (plus you actually didn't have the build that has a special cartoon mode and which most certainly helped XviD perform so well in the Futurama test). And nobody is going to stream a full D1 res movie at 200kbit/s, cartoon or life action (you're welcome to point out any errors in that statement by providing a link to an actual stream created by a respectable source which has those properties).
And since most cartoons I have seen are actually encoded in 4:3 you'd actually have to resize to 640x480 to retain the proper resolution... in my case you're saying that I should've used 640x480 rather than 608x464... I already gave you the bits/pixel*frame ratio above... c't did a test using those bits/pixel*frame ranges and the author of that test agrees that such low ratios are not in the range of a normal DVDrip.

[Satirik]

there is no normal dvdripp its not you who make comparison with basic parameters its the people who encode dvd who look your test and who adapt there params ... and thats why no one try to encode dvd anime with bitrate lower than 800 whereas encoding the same dvd in RV9 with a bitrate of 600 give the same quality (or very few less) ... and encoding the same with 400 will do an out file without less detail if you watch it 3 meters far from your screen ...

[Sirber]

The one in RV9 is 8/10, and the one on XviD is barely watchable, even with PP, 3/10.

[duartix]

Since someone has issued the cartoon codec comparison, I'm taking this ride just to post a very small critic here.
From looking at the DVD source (even though a very clean one acording to doom9) there seems to be some minor but clear artifacts.
Here we have two scenarios:

1 - Those are JPEG artifacts. I don't have any critic at the test.
2 - Those are MPEG2 artifacts. If this is the case then the source should have been filtered with a 2d cleaner which would definately would give a cleaner image for the codec to work than the plain source.

I guess only doom9 can answer this.

[Satirik]

Originally posted by duartix
Since someone has issued the cartoon codec comparison, I'm taking this ride just to post a very small critic here.
From looking at the DVD source (even though a very clean one acording to doom9) there seems to be some minor but clear artifacts.
Here we have two scenarios:

1 - Those are JPEG artifacts. I don't have any critic at the test.
2 - Those are MPEG2 artifacts. If this is the case then the source should have been filtered with a 2d cleaner which would definately would give a cleaner image for the codec to work than the plain source.

I guess only doom9 can answer this.

stupid critic not allowed here ;)

[Doom9]

it's dead easy.. if it's in the original shot (which is present, it's always the first shot) it's a JPEG compression artifact, if it's not, it's a compression artifact. I redid all the Futurama shots using a much higher quality setting and they should be virtually artifact free (but are also more than twice the size).
And you're driving down a dangerous road when you suggest filters to me. I don't test filters, I test how codecs can handle DVD sources.

[Atamido]

@Doom9: Is there any possibility of getting a .Torrent download of a RAR of all of the original PNG's of the screenshots?

[Doom9]

I hardly know what bittorrent is.. but if anybody would like to set that up I could send him/her the rared originals (uncompressed bmp images.. roughly 30mb).

[duartix]

it's dead easy.. if it's in the original shotYes it is.
it's a JPEG compression artifact, if it's not, it's a compression artifactI can't see the connection. If it was in the MPEG wouldn't it be also in the original shot?
Well I guess I can only draw my conclusions from looking at the uncompressed sources.And you're driving down a dangerous road when you suggest filters to me. I don't test filters, I test how codecs can handle DVD sources.I understand and respect your methodology completely. Using a filter is like adding another variable to the test, and what a variable it is...
Filters can interact with codecs in many unpredicted ways.

My sugestion was in the direction that most "anime freaks" follow, which is to pre-filter the source and make it even cleaner than the original.

You can't escape artifacts, but you can escape artifacts^2.

[Doom9]

@duartix: I compared the bmp of the original and the compressed shots.. there were no artifacts in the uncompressed original.

[duartix]

Thanks doom9!
:)