Quantcast FFmpeg: Comparison: Mencode-Me vs QuEnc vs TMPGEnc - digitalFAQ.com Forums [Archives]
  #1  
04-25-2004, 05:56 PM
audioslave audioslave is offline
Free Member
 
Join Date: Mar 2003
Location: Sweden
Posts: 725
Thanks: 0
Thanked 0 Times in 0 Posts
@all
I felt I had to tell you what I discovered a few minutes ago. I did some test encodes of a clip from Gothika on three different encoders: Mencode-me 0.21, QuEnc 0.47 and TMPGEnc 2.521.

Issue #1:
The samples encoded with Mencode-me and QuEnc have screwed up colors. The samples have a blue tone to them. The sample encoded with TMPGEnc looks like the source (no, not in quality ) - no color screw-up there.

Issue #2:
The sample encoded with Mencode-me are a few frames ahead when analyzing the samples in VDub This would make audio out of sync when muxing...

Issue #3:
Quality wise, QuEnc has the best looking sample - except for the color screw-up - with less artefacts (gibbs) than the other two samples.

Screenshots here:
http://hea.port5.com/Comparisons/Comparisons.htm

P.S. I forgot to mention all three samples are MPEG-2. I case you wonder...
__________________
AudioSlave
Reply With Quote
Someday, 12:01 PM
admin's Avatar
Site Staff / Ad Manager
 
Join Date: Dec 2002
Posts: 42
Thanks: ∞
Thanked 42 Times in 42 Posts
  #2  
04-25-2004, 06:30 PM
bigggt bigggt is offline
Free Member
 
Join Date: Mar 2003
Location: IamCanadian
Posts: 848
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by Audioslave
Issue #2:
The sample encoded with Mencode-me are a few frames ahead when analyzing the samples in VDub This would make audio out of sync when muxing...
Did you check force film because it has been determined that this is throwing things off

BTW glad to see you testing buddy
Reply With Quote
  #3  
04-25-2004, 06:37 PM
audioslave audioslave is offline
Free Member
 
Join Date: Mar 2003
Location: Sweden
Posts: 725
Thanks: 0
Thanked 0 Times in 0 Posts
@bigggt

Quote:
Originally Posted by bigggt
Did you check force film because it has been determined that this is throwing things off
Yes, Forced Film was used . I have read a few other posts were members have had difficulties with the audio sync when the movie was encoded with MEncoder. But then again, maybe they forgot to use Forced Film...

Quote:
Originally Posted by bigggt
BTW glad to see you testing buddy
Thank you .

BTW Did you check out the screenshots? I think it's wierd that the colors of the video can differ so much between the encoders.
__________________
AudioSlave
Reply With Quote
  #4  
04-25-2004, 06:46 PM
bigggt bigggt is offline
Free Member
 
Join Date: Mar 2003
Location: IamCanadian
Posts: 848
Thanks: 0
Thanked 0 Times in 0 Posts
Yes i looked at the screenshots and there is a quite noticable colour difference

It is weird but nothing suprises me anymore
Reply With Quote
  #5  
04-25-2004, 07:00 PM
Dano Dano is offline
Free Member
 
Join Date: Apr 2003
Location: Brockton, MA
Posts: 134
Thanks: 0
Thanked 0 Times in 0 Posts
I would have to agree with your conclusion about QuEnc, plus it is a little faster than Mencoder. I haven't seen the color problems though. It is a shame the libavcodec's mpeg-1 is no good.
__________________
-Dano
Reply With Quote
  #6  
04-25-2004, 07:20 PM
audioslave audioslave is offline
Free Member
 
Join Date: Mar 2003
Location: Sweden
Posts: 725
Thanks: 0
Thanked 0 Times in 0 Posts
@Dano

Quote:
Originally Posted by Dano
It is a shame the libavcodec's mpeg-1 is no good.
It certainly is . But, if the color problem is correctable (I'm sure it is ) I can live with MPEG-2 encodings when they looks as good as this .
__________________
AudioSlave
Reply With Quote
  #7  
04-25-2004, 07:40 PM
jorel jorel is offline
Invalid Email / Banned / Spammer
 
Join Date: Aug 2002
Location: Brasil - MG - third stone from the sun
Posts: 5,570
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by audioslave
But, if the color problem is correctable (I'm sure it is ) I can live with MPEG-2 encodings when they looks as good as this .
me too audioslave.
i like QuENC too much...and little corrections will come..but it's very good!

your pictures and tests are cool!
thanks for that.
Reply With Quote
  #8  
04-25-2004, 07:51 PM
audioslave audioslave is offline
Free Member
 
Join Date: Mar 2003
Location: Sweden
Posts: 725
Thanks: 0
Thanked 0 Times in 0 Posts
@jorel
Quote:
Originally Posted by jorel
your pictures and tests are cool!
thanks for that.
Thank you .
__________________
AudioSlave
Reply With Quote
  #9  
04-25-2004, 08:29 PM
Prodater64 Prodater64 is offline
Free Member
 
Join Date: Mar 2003
Location: Palma de Mallorca - España
Posts: 2,925
Thanks: 0
Thanked 0 Times in 0 Posts
But if before of mencoder boom, many people said that it was and is better than tmpgenc in mpeg1 and mpeg2 too. With much tests, like audioslave one.
How do you test mecoder. With make avis, internal filter, how?

Audioslave, check your results with other posted in the forum.

See you.
Reply With Quote
  #10  
04-25-2004, 08:45 PM
vmesquita vmesquita is offline
Invalid Email / Banned / Spammer
 
Join Date: May 2003
Posts: 3,726
Thanks: 0
Thanked 0 Times in 0 Posts
@audioslave
This test result is wierd, because MencodeMe should produce an output better than QuEnc since it's using parameters bilu, digital.doc and myself took weeks to optimize, while QuEnc doesn't allow much customization. Actually they did most of the work, I just did the final tests. Anyway, you didn't post what settings were used in your tests in MEncodeME.

Plus: as posted yesterday (so you probably didn't knew this), forced film should not be used in progressive MPEG2. You should leave field operation in none (update to 0.22) and use 23.976 fps as framerate.

Please post the conditions of your test, if you used internal filters, optimzal script, etc.

EDIT:I've never had this color problem, but it can be fixed using mencoder internal filters or avisynth.
EDIT2:if possible, post small clips. Pictures don't say too much, because a encoder may have an I-Frame in that specific frame, while the other choose to put a lower quality B-frame, for instance,
Reply With Quote
  #11  
04-26-2004, 04:52 AM
rds_correia rds_correia is offline
Free Member
 
Join Date: Apr 2003
Location: Chinese Democracy starts now!
Posts: 2,563
Thanks: 1
Thanked 0 Times in 0 Posts
Hya all,
In fact when running comparison tests I would personally use VOB sources without any PP filtering on any encoder
After that I would try our dayly technics such as MA script,etc.
That will give us a more clear picture of what the encoder really is capable of doing on it's own.
Anyway great to see that day by day we have more people testing mencoder and other encoders
Cheers all
__________________
Rui
Reply With Quote
  #12  
04-26-2004, 05:44 AM
audioslave audioslave is offline
Free Member
 
Join Date: Mar 2003
Location: Sweden
Posts: 725
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by vmesquita
Plus: as posted yesterday (so you probably didn't knew this), forced film should not be used in progressive MPEG2. You should leave field operation in none (update to 0.22) and use 23.976 fps as framerate.
That I didn't know . I will make a new test as soon as I get the time.

Quote:
Originally Posted by vmesquita
Please post the conditions of your test, if you used internal filters, optimzal script, etc.
I will, again, as soon as I can .

Quote:
Originally Posted by vmesquita
if possible, post small clips. Pictures don't say too much, because a encoder may have an I-Frame in that specific frame, while the other choose to put a lower quality B-frame, for instance
I would post clips if only I had the space. Right now I don't have that possibility - sorry...

Quote:
Originally Posted by rds_correia
In fact when running comparison tests I would personally use VOB sources without any PP filtering on any encoder
Yes, that might be true, but since I'm always using scripts for my encodes I did the test with a script. But I get your point .

@vmesquita, bilu & digital.doc
I hope you didn't get offended by my test? That was not my intention. Don't shoot me, I'm just the messenger .

@All
I just thought it would be a good idea to post my findings here so the developers of these great encoders and GUI's can improve them even more !
__________________
AudioSlave
Reply With Quote
  #13  
04-26-2004, 05:56 AM
incredible incredible is offline
Free Member
 
Join Date: May 2003
Location: Germany
Posts: 3,189
Thanks: 0
Thanked 0 Times in 0 Posts
Send a message via ICQ to incredible
@ Audioslave

What was the size of your source and the encoding and ... which avg Bitrate was used????

1. I do assume that the colors of Qenc and MencodeMe are CORRECT! And the one from TmpgEnc are not, why? Very simple ....: TmpgEnc is the only one which does internally process an out of YUV chroma conversion -> to RGB24! Qenc and MencoderME (both libavcodec) do keep the YUV subsampling means Y/C separation. so a conversion from YV12 to YUY2 (if done internally by libavcodec as I assume) will be not that worse to colors if a conversion to a RGB color model would be preformed like TmpgEnc does.

2. I see clearly that Qenc here is the winner! (but to be seen on that sample! So Vmesquita is right that that cant be certified for general!)
How do I see that: Very easy I downloaded your samples using the sRGB (windows) colorprofile and used Photoshop to see a sRGB colorproof.
I opened a new sRGB color document where I imported your 3 shots of the SAME frame as layers. After that I generated very above a "gradation setting layer" which affects all your pics to be more light!

If you watch (at 200%) the wall and the details and so on by switching the image layers on and off... Qenc gave the smallest Quantisation artefacts , means almost none.

Here is a Sample (a photoshop document)
DO NOT DOWNLOAD IT IF YOU DO NOT GOT PHOTOSHOP 6 or above which supports setting-layers! As if not, you just stretch my traffic on my free webspace for nothing .

http://home.arcor.de/the.incredible/compare.psd

(BTW Audioslave, your MencoderME example was encoded in another proportion then the other two! Means scaled larger within the final resolution )

Here you can see the images separately... for those who dont got Photoshop:

Qenc


MencodeMe


TmpgEnc
Reply With Quote
  #14  
04-26-2004, 06:39 AM
audioslave audioslave is offline
Free Member
 
Join Date: Mar 2003
Location: Sweden
Posts: 725
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by incredible
1. I do assume that the colors of Qenc and MencodeMe are CORRECT! And the one from TmpgEnc are not, why? Very simple ....: TmpgEnc is the only one which does internally process an out of YUV chroma conversion -> to RGB24! Qenc and MencoderME (both libavcodec) do keep the YUV subsampling means Y/C separation. so a conversion from YV12 to YUY2 (if done internally by libavcodec as I assume) will be not that worse to colors if a conversion to a RGB color model would be preformed like TmpgEnc does.
I'm sorry but I don't agree with you on this one. The clip from TMPGEnc is the only clip of the three that looks like it should - color wise! Download this *.psd and see for yourself.
http://hea.port5.com/Comparisons/EncComp.psd
I've put the same frame, from all encoders + the un-encoded script, in layers. It cleary proves my previous statement .
@All
As incredible said: If you don't have Photoshop don't bother with the download .

Quote:
Originally Posted by incredible
(BTW Audioslave, your MencoderME example was encoded in another proportion then the other two! Means scaled larger within the final resolution )
No it's the same as the other two. I see that you used frame #1341 from my Mencode-me screenshots on my comparison page. Frame #1341 from the Mencode-me clip is actually 2 frames ahead when compared to QuEnc and TMPGenc . See my previous posts.

@All
Comparison page updated:
http://hea.port5.com/Comparisons/Comparisons.htm
Added info on what settings I used for each encoder + the script I used.
__________________
AudioSlave
Reply With Quote
  #15  
04-26-2004, 06:57 AM
Abond Abond is offline
Free Member
 
Join Date: Mar 2003
Posts: 243
Thanks: 0
Thanked 0 Times in 0 Posts
Good comparison. The problem as I see it is you are using extreme settings for QuEnc, but not for the others (for example CQ 100 for TMPG, High Quality, for mencoder I don't know)
Reply With Quote
  #16  
04-26-2004, 07:02 AM
FuPP FuPP is offline
Free Member
 
Join Date: Mar 2004
Posts: 92
Thanks: 0
Thanked 0 Times in 0 Posts
Hi man !

I have PM you a few remarks about your script . Btw, I think that for such a comparison, a more neutral script would be far better (no filtering at all I mean)

FuPP
Reply With Quote
  #17  
04-26-2004, 07:58 AM
vmesquita vmesquita is offline
Invalid Email / Banned / Spammer
 
Join Date: May 2003
Posts: 3,726
Thanks: 0
Thanked 0 Times in 0 Posts
Quote:
Originally Posted by audioslave
@vmesquita, bilu & digital.doc
I hope you didn't get offended by my test? That was not my intention. Don't shoot me, I'm just the messenger .
I was not doing that. I just think that comparissons that doesn't show it's conditions doesn't make sense. But its ok now that you put the settings used.
Reply With Quote
  #18  
04-27-2004, 05:03 PM
Jellygoose Jellygoose is offline
Free Member
 
Join Date: Jun 2002
Location: Germany
Posts: 1,288
Thanks: 0
Thanked 0 Times in 0 Posts
Just wanted to share my experiences.
I tried a whole encode today using QuEnc 0.47, with Trellis, VBR, High Quality, 2-pass checked and also the "Extreme and Slow" option enabled. The movie I encoded was Charlie's Angels 2. All I can say is, the results were EXTREMELY BAD.
Gibbs and blocks all over the place. High Action scenes, were full of blocks, scenes with many details/colors had artifacts around objects. Low lit areas looked very unstable, with dancing DCT-blocks, ALTHOUGH i already used a blockbuster line with variance of 0.5! Here's the complete script:

Code:
Crop(8, 72, 704, 432) # 72_72 
Deen()
Undot() 
Unfilter(10,10)
DCTFilter(1,1,1,1,1,1,0.5,0) 
Blockbuster(method="noise",detail_min=1,detail_max=8,variance=0.5,seed=4326)
Addborders(0,72,0,72) # 72_72
On top of that I had these "inverted blocks" issue in several scenes. btw: Average Bitrate is 2525kb/sec, MPEG-2 704x576.

The exact same encode with Mencoder came out almost perfect. Well let's say perfect compared to the QuEnc result. Gibbs and DCT blocks were gone, and blocks in high action scenes were less obvious. In fact the max. bitrate was limited to 8000kb/sec. I figure setting it to 9800kb/sec will make this problem obsolete. Sorry to say that, my choice for MPEG-2 encodings will be Mencoder just now...
__________________
j3llyG0053
Reply With Quote
Reply




Similar Threads
Thread Thread Starter Forum Replies Last Post
Quenc x mencode - Vmesquita estranho74 Conversão e Codificação de Vídeo (Português) 8 07-16-2004 01:03 PM
FFMPEG: Another QuEnc 0.51 Error darkeye Video Encoding and Conversion 2 05-18-2004 01:34 PM
FFMPEG: QuEnc 0.46 is out Abond Video Encoding and Conversion 12 05-04-2004 06:41 AM
FFMPEG: Min/Max Bitrate in QuEnc? Icarus3000 Video Encoding and Conversion 13 04-28-2004 10:07 PM
FFMPEG: QuEnc GFR Video Encoding and Conversion 78 03-28-2004 04:05 PM

Thread Tools



 
All times are GMT -5. The time now is 05:47 PM  —  vBulletin © Jelsoft Enterprises Ltd