CCE vs QuEnc vs imToo [comparison]

[Borat Sagdiyev]

CCE 1 pass @ 1000
http://img282.echo.cx/img282/2202/frame76cce271pass10006dz.jpg

imToo 1 pass @ 955
http://img282.echo.cx/img282/815/frame76imtoo1pass9556cx.jpg

QuEnc 1 pass @ 955
http://img282.echo.cx/img282/2427/frame76quenc1pass9559al.jpg

QuEnc 2 pass @ 955
http://img282.echo.cx/img282/4088/frame76quenc2pass9551yu.jpg

QuEnc 1 pass @ 955 with HQ
http://img282.echo.cx/img282/3829/frame76quenchq1pass9557vg.jpg

[kolak]

That's not any test:( You can find better tests on doom9 forum.
Subject CCE vs. xxxx encoder I think is closed.
Everyone has his "the best " encoder:)

[Borat Sagdiyev]

I posted this because the difference between imToo (30 dollar) and CCE is striking.

[Encoder Master]

I tested also imTOO but the quality is very bad for me. The best choise in your case is realy QuEnc 2pass and high quality because it's the best quality and you don't have to pay money.
Can you encode with HC Encoder with the same sample? But load this samples first in VDubMod to be sure that you always use the same frame type like p- or B-Frames.

[kolak]

Look here http://images.industryclick.com/files/127/404VSsfeature.pdf
This test was made by profesionalist.
I test encoder by waching clip in Powerdvd in 4 x zoom mode, not by compareing still frame.

[dragongodz]

Borat Sagdiyev - just posting some pictures of individual frames says nothing. you should say what you see with actual playback. you should also try CCE 2 pass.

i also note the frame size is SVCD(480x576), 1000 and 955 are pretty low for that. no encoder is going to look brilliant and single frames become even more questionable. so a short clip would say much more than showing those.

[Fishman0919]

You can't judge encoders by a still pic... you might be getting an I frame from one and a B frame from another... and I've said it before... Judging encoders by still pic is like judging the taste Ice cream by looking at it it's just wrong and silly... one frame can't tell you how good the next frame is in relation to the one before or after it. The only way to judge video is to watch it.

Edit: lol dragongodz... you beat me to it

[Encoder Master]

I think you should got with ImToo the same Quality as with QuEnc because it runs also with the avcodec.

[unplugged]

Just put a stupid "TemporalSoften(4,2,2,15,2)" in your script you'll see that CCE and HC fly to best quality almost all times.

Libavcodec (QuEnc, FreeEnc...) encoders has some integrated adaptive masking really crappy for my taste, have no care for details!

No thanks, I have a more efficient way to remove noise from detail:
AVISYNTH.

[Encoder Master]

Libavcodec (QuEnc, FreeEnc...) encoders has some integrated adaptive masking really crappy for my taste, have no care for details!

the avcodec also removes details because it uses a H.263 quantizer and not a MPEG-quantizer.

[Nic]

@unplugged: Can I see the evidence of QuEnc using "integrated adaptive masking" before you call it cr*ppy next time? Although many forms of masking can be applied in libavcodec. QuEnc does not apply any.

@EncoderMaster: QuEnc uses the correct MPEG matrix as defined in the MPEG-2 Spec ISO 13818-2 part 6.3.11. But of course you can load any matrix you like into QuEnc for encoding.
(EDIT: oh, I may have misunderstood, you don't actually mean the matrix used)

[Encoder Master]

(EDIT: oh, I may have misunderstood, you don't actually mean the matrix used)

Yes, I don't mean the matrix. This is also why QuEnc or other avcodec based encoder give such a good quality on low bitrates. The quantizer is lower and smother then the normal MPEG-standart. But I fI want to use the MPEG-quantizer then it doesen't work with avcodec for example in MEncoder oder FreeEnc.

@Nic, again.

Can you tell me what settings improve the quality if you set the Extreme and slow settings? So I want to test it also with a ini file in FreeEnc. Hope you can help me out with the settings you compiled QuEnc.

[Inc]

Yes, I don't mean the matrix. This is also why QuEnc or other avcodec based encoder give such a good quality on low bitrates. The quantizer is lower and smother then the normal MPEG-standart.
You mean the Quantizer Value/Curve result from Bitrateviewer?
Thats cause Libavcodec uses a linear quantisation where CCE TmpgEnc etc. use a non-linear quantisation. So at very dark underwaterscenes a Quant Value of 2 already results in some blocks if you look closely. Thats why I choosed in my mencoder lines a min Quant of 1 using no trellis.

[kolak]

Again unneeded discussion about CCE vs. xxxx encoder. There was so many discussions like this....

To Borat Sagdiyev:
CCE SP costs about 2K$ but it's not for home user. It's dedicated rather for post-production studios( for them 2K$ is not a big money). I can say that CCE is very good encoder e.g in comparison to Sonic SD2000 hardware encoder which costs 20K$!! I worked on both and I made many comparisons between them. CCE almost always was better and it has excellent support (best regards to Ray Wallick) when SD2000 has many bugs in software part:(
For home user CinemaCraft has basic version for 60$ with excellent quality and speed.
We should be happy that there are peoples who make very good and free encoders. Many thanks for them!

Andrew

[unplugged]

Originally posted by Nic
@unplugged: Can I see the evidence of QuEnc using "integrated adaptive masking" before you call it cr*ppy next time? Although many forms of masking can be applied in libavcodec. QuEnc does not apply any.
I have no reasons to do this so logically I don't want offend you nor ffmpeg developers, I simply can't understand the enthusiasm for the mpeg2 encoder inside, that said for comparison I consider ffdshow a multimedia monolith with no doubt and at same time the ff mpeg2 encoder just a good encoder (despite years have passed and projects like XviD in the meanwhile have quasi *warped* its quality).

I agree there can be argue for certain cases or with interlaced encoding but...
With ~3000 kbit/s progressive usual scenario QuEnc (and mencoder, freeenc...) tends to flat or snub the importance of certain details/areas of image in way that I have already seen (as a "mark") with H.264 encoders.
This (bad) effect performs like an aggressive/ONEWAY temporal threshold-based filter applied systematically, certain parts becomes blurry or glued between adjacent frames without no worry by the encoder.

Plus the quantization over all the image or frames gets an inconsistent fashion as opposite as I see with CCE or HC (or with the great supertuned XviD), you just need a 2x view in your VirtualDub session to see how much blocks are spreaded in moving areas.

Crucial point:
Keep in mind that 99% of DVD movies are *really* progressive even when they are stored as interlaced, once the source is clean itself or cleaned by pre-denoising and encoded in progressive mode there is no story: CCE and HC (especially the last one for mid-low bitrate range) have an unbeatable *consistency* with details, no blur no blocks.

I suspect that the majority of comparisons that give the edge to QuEnc in comparison of CCE/HC are based on interlaced encodings, and many people doesn't even know that it ISN'T WORTH (at all bitwise) to compress a film (visually progressive) in interlaced mode just because it was originally stored in that STUPID MISLEADING WAY.
(with 99% of cases there is NOT A SINGLE combed/interlaced line!)
Do interlaced the really interlaced content, so do that way the extras! ;)

[kolak]

Keep in mind that 99% of DVD movies are *really* progressive even when they are stored as interlaced, once the source is clean itself or cleaned by pre-denoising and encoded in progressive mode there is no story: CCE and HC (especially the last one for mid-low bitrate range) have an unbeatable *consistency* with details, no blur no blocks

That's true. If you have good source you can use even middle range encoder and you'll get good result.

[dragongodz]

or with the great supertuned XviD
you do realise the latest QuEnc is using a version of Xvid rate control ?

we are looking at improving QuEnc all the time and any USEFUL and CONCISE feedback on any problem areas may help to do that. saying you think encoder X is better alone helps nobody and doesnt encourage us to bother continuing at all.

Can you tell me what settings improve the quality if you set the Extreme and slow settings? So I want to test it also with a ini file in FreeEnc.
they are in the source code. dare i say if you cant find them or how they are used then you shouldnt be messing with them. ;)

[Encoder Master]

they are in the source code. dare i say if you cant find them or how they are used then you shouldnt be messing with them.

I know where they are but not to use them :( . So I ask and I never want to mess me.:) ;)