CCE 2.50 and 2.67 Image Quality Comparison Set

[Yuri]

CCE trial versions accidentally may spoil your encodings because of the truncate code and logo-encoding scheme. So BUY this wonderful piece of software to avoid any disagreeable surprises.

Gratitudes to jarvis1781 for his eclaircissements

One may still compare the CCE TRIAL versions spoilage at
http://sch-yuri.narod.ru/DVD/DVD.htm

[DDogg]

You did not mention your 2.67 quality tab settings. Without those, your statement is unsubstantiated. Please try these settings (http://forum.doom9.org/showthread.php?s=&threadid=64800) OR RB's Progressive template. (http://home.t-online.de/home/340044300675/ccetemplates.zip) Let us know how it went as it looks like you have some good comparisons set up.

[DDogg]

Yuri, sorry if I upset you. What I was getting at was did you uncheck all the settings in the CCE 2.67 Quality tab? Specifically, lowpass filter, Effect restricted vertical filter,and Dithered Quantization. This was not clear, at least to me, from your post.

Since you made such a bold statement in your post, I thought it best to try to get a slightly more clear set of data from you.

[DDogg]

Well it sounds more apples to apples. Since we are now together on the importance of the experiment's objectiveness, the filter chain scares me a little. Yep, sure the same input was fed to the encoders, but I am wondering if perhaps one version of CCE reacts differently to the shapening, resizing ?

It might further prove you point if a stripped down script was used with just the bare essentials with the least amount of filtering possible. This would give even a better objective test (I think it would at least). Something like:

Mpeg2Source("VTS__01_P01.P.16~9_1.d2v")

ConvertToYUY2(interlaced=true) #??? Needed?
Telecide (0)
Decimate(cycle=5)
AssumeFps(25)

#might be interesting to use simple resize? As far as I know it
#effects the video the least of any resizing filter.
# might even want to no-op the resizing for the test???

LanczosResize(720,576)
LanczosResize(720,368)
Crop(0,0,720,368).AddBorders(0,104,0,104)

Anyway, just a few thoughts for you to consider. This is an important test. If somebody else can then replicate the same results we will have a more clear method of judgement. Do you have the bandwidth available to place a small snip for download?

[Don R.]

Does anyone have an opinion why the older versions of SP provide better quality at the higher bitrates than the new version?

I keep noticing a haze over SP encoded MPEG-2 files used for creation of DVD's.

[RB]

Originally posted by Yuri
The only filter that can't be disabled either in CCE 2.50 or CCE 2.67 is the Image Quality Priority - so I set it to 15 both in CCE 2.50 and CCE 2.67 this is given on my comparison page in the line "The CCE 2.50 and 2.67 Settings: One Pass VBR, bitrate=2000-9000, Q=20, IQP=15, Intra DC=10, AR=4:3, Progressive."
The range for IQP in CCE 2.50 is 0-100 whereas in CCE 2.67 it is 0-64. So you should have set it to 10 in CCE 2.67. Generally, CCE 2.67 should give better quality with a higher IQP. Try setting it to 34-40.

[Don R.]

Frame #44 and #45 illustrate my subjective conclusion concerning 267 and that is SP lacks contrast and appears to have a fog across the image. #44 may actually have higher resolution than #45, but it is so "flat" that the higher resolution is lost.

#45 has greater contrast and APPARENT resolution and is more appealing to my eyes.

[chilledoutuk]

OK so from your tests 2.50 is better quality than 2.67 but i thought all 2.67 versions were betas and thus is not a fair test.

I suggest the test is rerun with 2.50 vs 2.66.01.07 (or the latest non beta which i think that is)

This way we will see if it is a bug in cce 2.67 that is causing this significant amount of artifacting.

[Don R.]

If 2.67 is not as good as 2.66 or 2.50, where does one download either 2.50 or 2.66.01.01?


The latest beta is 2.67.00.22 that CCE says is "revised to support Adobe Premiere.

It has been nothing but beta, beta, beta on the Cinema Craft website for months. Can't Cinema Craft get it right?

Another question that must be on eveyone's mind. If MainConcept 1.4 and Sorenson Squeeze provide more pleasing MPEG-2 files, why on earth pay 10X more for Cinema Craft SP?

[Steve56]

I did some tests with CCE 2.67.00.13 and what I found is that when you only encode one pass, then some frames are really "blocky".

When rendering multipass at least (vaf + 1 more pass) the output gets much better!

Don R.:

MainConcept 1.4 is really bad against CCE, just a hole class!

Sorenson Squeeze does not produce MPEG2 but MPEG4 and their own Sorenson Codec, they are both very good but not MPEG2!

Steve56

[Don R.]

My method of encoding with CCE-SP is using the 1st VBR analysis pass and then change to multiple VBR passes and select four passes.

So far as I can determine, more than four VBR passes don't improve quality.

[Don R.]

Steve, Sorenson Squeeze 3.5 does encode MPEG-2 and it uses the MainConcept MPEG-2 technology.

I have found it even does a great job on one pass CBR.

[yup]

Yuri try CCE 2.64 and linear Quant Scale. And use for CCE prefilter in Avisynth script, CCE only encoding not as TMPGEnc or Canopus Procoder. I use for DVD to CVD conversion only CCE 2.64 begining from 2.66 linear quant scale obsolete.
yup.

[MLS]

Read the FAQ, linear quant is not for MPEG2.

/MLS

[DDogg]

Yuri, :thanks: Nice job!

[wmansir]

Yuri, could you post the BitrateViewer shots for the multipass tests too? I would like to see how the bitrate was allocated.


It looks like the one-pass VBR test was messed up for 2.67, not by you, but by the encoder. I'm curious what was the difference in file size of the output? It appears they both start and end at the same bitrate, but 2.67 dips way down in the middle (to the minimum 2000kbps), where 2.50 stays relatively flat. Which would lead me to believe the 2.76 output would be smaller. Perhaps the Q values were changed between builds like the Image Quality settings were, so Q20 in 2.50 does not equal Q20 in 2.67? I never use 1-pass VBR so I wouldn't know.

[br408408]

I'm wondering what 2.66 does as it runs quite a bit faster than 2.5 on my P4

Bill

[troy]

Wow thanks. I wonder why it has taken this long for someone to do this kind of comparison. I would love for someone else to do this and see what they come up with. I am too much of a nubie to do this but I have suspected this as my cce 2.5 encodes were always good and it didn't make any sense to think cce 2.66 was worse but it did seem so.

[chipvideo]

It also seems to me that the 2.67 version is optimized for hyperthreading. Also it sounds like 2.50 runs faster on amd systems. Looks like I will just stay with 2.50.

3.15RT encodes is good enough for now with bicubic resizing dvd's.

[onesoul]

@Yuri

What computer did you use on your tests?

[Boulder]

Originally posted by br408408
I'm wondering what 2.66 does as it runs quite a bit faster than 2.5 on my P4

Bill

At least Cinema Craft is optimistic:


Version 2.66.01.01 Friday, September 13, 2002

Changes
- Picture quality has improved.
- Filter setting is radically changed.
- Added inverse 3:2 pulldown function.
- Supported sequencial still image files.
- Supported Microsoft DV Codec.
- Enabled to set chapter by timecode.
- Enabled to playback movie on chapter setting window.
- Plug-in has almost as same features as that of standalone version.
- Enabled to start up from batch file.
- Added context sensitive help.

[Don R.]

Why would Cinema Craft shoot itself in the foot by placing on the market software that is worse than an earlier version?

Cinema Craft must be fighting against losing market share to Sorenson, MainConcept, and TMPGENc not to mention the hardware encoders that are coming down in price.

What would be Cinema Craft's strategy?

[troy]

Just a totally nubie thought. Maybe the Q settings are not on a linear scale. Maybe 12 does not equal 20.

[mb1]

@ Yuri

In my opinion there are some 'false' settings. but first let me say I only tested 2.50, 2.64 and 2.66 myself. 2.66 always gave the best picture. 2.67 is in beta stadium (for a long time now).

1. significant to me is the different gop used by the two versions. CCE changes its set gop only on scene changes :rolleyes: ...

2. very short test clips (70 and 100 pal frames) with very different bitrates f.e. first test: cce 2.50 used avg 5801 whereas cce 2.67 used avg 4745

3. you did set avg 6000 in multipass mode. but the resulting avg bitrates are way below (cce 2.50: 3624 kbps, cce 2.67: 3968 kbps)

4. if I look at your listed gops (from BV) there are differences with those listed for the pictures

5. example: you did compare frame 44 (P-Type) with frame 45 (P-Type, but isn't; in my opinion you showed frame 45 which is a B-Type, whereas frame 44 would be a P-Type)

I'm sure there are more 'irregularities'.

Also please do tests with at least 10 seconds of material, 20 would be even better.

[sh0dan]

Interesting result! Thanks for the detailed info.

It would seem like CinemaCraft has changed their matrix, resulting in higher quantizers, or they have introduced an ME-bug.
Is there any way of comparing the two matrices? (I cannot remember if it can be viewed in 2.50).

Would you mind trying out an a larger source? (like 1000+ frames)

[Don R.]

Is it the consensus that SP users would secure better image results if they used the old 2.66 version of SP rather than replacing it with the beta 2.67?

[Don R.]

Have you tried reducing bitrate?


Also, which verision of 2.66 would you recommend?

[Matthew]

I've encoded quite a bit of material using 2.67.00.11 so now I feel depressed.

I suppose I shall go back to the latest build of 2.66, which is 2.66.01.07.

Blah.

[Don R.]

Thank you for the recommendation.

I am still studying the hardware and software encoders.

A pretty darn good source told me BASIC may well be as good as SP.

Maybe we ought to be comparing BASIC to SP and the other hardware-software encoders.

[Matthew]

There are comparison tables for basic/SP

http://www.cinemacraft.com/eng/ccebasic.html
http://www.cinemacraft.com/eng/ccesp.html

AFAIK basic has the same encoding engine, just there are restrictions (only form of VBR is 2 pass). So it is as good as SP if you only ever use 2 pass VBR (or CBR with 2 passes or less). Otherwise, it's inferior - as you'd expect given the price :)

[troy]

I want to do some tests on some interlaced dv video. What is the best way to compare m2v files and get a good screenshot. I can not open the video and get good screenshots with mediaplayer. I guess I could convert them to dvd and play them on my computer with windvd but I was interested to know if there is a better way to do this.

[Boulder]

Originally posted by troy
I want to do some tests on some interlaced dv video. What is the best way to compare m2v files and get a good screenshot. I can not open the video and get good screenshots with mediaplayer. I guess I could convert them to dvd and play them on my computer with windvd but I was interested to know if there is a better way to do this.

You can load the files in VirtualDubMod and take snapshots of the frames you wish.

[wmansir]

Before everyone burns their 2.67 backups...

I just did a few quick tests with 2.5, 2.66 and 2.67. Unfortunately, I don't have any webspace to upload screen shots to. But my results were no where near as dramatic as Yuri's. In fact I probably couldn't use .jpg to display the screenshots because the differences were that subtle.

My test source was Chapter 2 from Saving Private Ryan, the beach landing. I was looking for something that would test the encoders at average bitrates. I clipped two short 3-4 second scenes together from the part where Hanks was looking around at all the chaos (the best one was the flamethrower exploding). I used settings as close to Yuri's as possible (no filters, Q20 for VBR, 2k-6k-9k for Multi, Image priory 20 for 2.5, 12 for 2.66/7, ect). I did Multi and one-pass VBR for all versions.

There were a couple of very minor differences when examining the stills. I would say 2.5 had slightly more detail and mosquito noise compared to 2.66/7, but it was very very slight. There was no way I could spot the difference while the clips were playing. And the vast majority of the stills looked virtually identical.

Here are the Bitrate Viewer Numbers:

2.5
Size: 6,414 KB
Bitrate: 0/6994/8519
Quant; 0/4.79/6.14

2.66
Size: 6,488 KB
Bitrate: 0/7075/8534
Quant: 0/5.14/6.64

2.67
Size: 6,491 KB
Bitrate: 0/7080/8575
Quant: 0/4.92/6.42

So, it looks like in this case 2.5 did slightly better, and 2.66 slightly worse than 2.67, according to the average quant. 2.5's output file size was slightly smaller, but only around 1%. So, at least in my case it appears the Q values for 1-Pass VBR are roughly equal between versions.

Perhaps the issue is the source. I don't have any anime handy to test out. But in my test I did see any major differences.

[Boulder]

Like some other folks, I also suspect that the biggest problem when comparing 2.50 to later versions is the IQP setting, being from 0-100 and 0-64. Since no one really knows whether they can be directly compared, there's no way to decide which version does a better job:D

[Boulder]

Apparently 2.50 and 2.66 use matrices somehow differently or the matrix gets patched to 2.50 incorrectly (I did some tests on the very same matrix on both setups).

However, that's all I can say about that for obvious reasons:devil:

[kru]

I see that theese test are made for DVD bitrates, but is CCE 2.50 also the best choise for SVCD/CVD (1200-2530 kbps)???

[Boulder]

Yuri,

have you made similar tests with non-cartoon/anime material? I was unable to make such clear difference between the different encoder versions. (I agree, 2.67 looks like puke in that test:D )

And I don't mean to bash your test, not me! I'll take my hat off to anyone who just takes the time and runs tests and reports them too:) The more sceptical you are, the better the explanations you get:D

[Matthew]

Yuri perhaps you should make your source clip of that cartoon available (asusming it's legal ;) ;)) so it can be tested (not by me, as I have no idea), but allowing other people to fiddle with the settings seems like a logical course of action.

[Matthew]

Download that at 0.5 k/sec (ouch)

Uploaded it elsewhere, this will be much faster.

http://members.tripod.co.uk/jefffahey88/sample/tarzan_49_tricky_frames.zip

Copy/paste that link in new window.

Unless I'm mistaken frame 44 is frame 27 on this sample, and hell it does look crap when re-encoded with 2.67's 1 pass VBR. 4 pass VBR produces quite a bit better result though.

[jarvis1781]

Originally posted by Yuri
I've been expecting the question converning the matrix stuff, 'cos it's yet the only one left that's changable but, alas, I tired differnet ones in 2.66 and 2.67 with no good effect whatsoever.
IT'S THE ENCODING ENGINE ITSELF!

ok this is where I have to step in. I have been around these forums for ages...and I never once registered to post anything because any information i needed was right there...

I registered 5 days ago because of this thread. It seems that CCE is being bashed a little because "IT'S THE ENCODING ENGINE ITSELF!". No, there is no problem with the encoding engine. The problem is you need to PURCHASE this software if you want it to work correctly. You admit to using different matrix settings for it and these are what allow you to get rid of the "water mark". They are also seem to be causing you trouble with CCE...my legitimate version works faster and the results are indeed better as the CCE site claims. I suggest you actually purchase this software before you decide to make any more claims about it. Sorry if this post upsets anyone but I just got tired of hearing people bash CCE when there is nothing wrong with it...the problem is with the pirates who cracked it.

[Matthew]

jarvis, I think that's a bit of an assumption to make.

Having said that you seem to be correct. When the logo is in the picture the quality is MUCH better. These are each close to 150k stills (copy and paste in new window).

http://members.tripod.co.uk/jefffahey88/cce/nologo.jpg
http://members.tripod.co.uk/jefffahey88/cce/logo.jpg

What I find odd is that this problem is not widespread...you'd think that something like this, resulting from an imperfect fix, would be much more apparent.

There is one upside of this...this sample is a wonderful test of the effectiveness of a fix...I'll certainly be holding onto it.

[jarvis1781]

Originally posted by Matthew
jarvis, I think that's a bit of an assumption to make.

Having said that you seem to be correct. When the logo is in the picture the quality is MUCH better. These are each close to 150k stills (copy and paste in new window).

http://members.tripod.co.uk/jefffahey88/cce/nologo.jpg
http://members.tripod.co.uk/jefffahey88/cce/logo.jpg

What I find odd is that this problem is not widespread...you'd think that something like this, resulting from an imperfect fix, would be much more apparent.

There is one upside of this...this sample is a wonderful test of the effectiveness of a fix...I'll certainly be holding onto it.

well there you have it...visual proof that CCE works GREAT when it is used legitimately...

i don't care what people do as long as they don't bash GREAT software when there is nothing wrong with the software. again the problem lies not in CCE but in the "hack job" the hackers did with removing the logo.

[Matthew]

As the choice is now between 2.66 and 2.50 (I think :p) and I don't have eagle eyes, could someone tell me their opinion of the stills here:

http://members.lycos.co.uk/jefffahey88/cce2/

They are frames 27 and 28 (44 and 45 on Yuri's site), done with his settings except 4 pass, min/max 0/9200 and average bitrate 6000 (which gives about same size as 1 pass Q20).

Yuri's 1 pass stills don't reflect well on 2.66 (wmansir seems to agree with his Saving Private Lynch test) but I'm not so sure about 4 pass (which is what I always use).

Incidentally, for 2.66 1-pass with and without logo is also there, and they look identical to me, so I don't think patches are an issue.

It is perhaps ridulous that so much is being read into a couple of frames, but whoever said humans were rational ;)

[Matthew]

Oh and Jarvis, to my eyes even with the logo 2.67 looks inferior to 2.50 for that still. However I am only referring to 1 pass here.

2.67 http://members.lycos.co.uk/jefffahey88/cce/logo.jpg
2.50 http://members.lycos.co.uk/jefffahey88/cce/2.50,nologo,1pass.jpg

Look at the thumb.

And btw, while the source was a CCE 2.50 clip from Yuri, the 2.50 image above was taken from a mpv outputted by my CCE 2.50 - not the source.

The difference might be explained by image quality priority, testing that is too much of a pain in the backside for me though.

[syzygytec]

A test by a dummy *me

http://www.geocities.com/syzygydvd/nemo.htm
If your really interested have a look but be warned it's on GeoCities and only has a transfer limit of 10mb per day so don't download just to download. File is 917k

3 screen grabs via WinDVD 5

dvd.bmp = original DVD

cce.bmp = Backup via D2S / FACAR / D2SRoBaOPV / CCE 2670013

shrink.bmp = Should be the same as dvd.bmp, DVD Shrink removed everything but main title and main soundtrack, = 0% compression so should be just a copy of the original

I see little to no difference

[Matthew]

AFAIK there are multiple checks done by the CCE trial and it's possible that one may have been missed I suppose...dunno...

Good point Yuri about the logo, I hadn't thought of that.

However, I just encoded the clip with and without logo at 9000 CBR and while frame 27 is actually a bit better without the logo (!), frame 28 displays image destruction only when the logo isn't there :(

And another test...for 4 pass VBR, 6000 bitrate, 2.67 with logo = 2.67 without logo. At least IMO. Here are stills:

http://members.lycos.co.uk/jefffahey88/cce3/2.67,4pass,logo.jpg
http://members.lycos.co.uk/jefffahey88/cce3/2.67,4pass,nologo.jpg

Give your honest opinion Yuri =)

[sh0dan]

Originally posted by Yuri
Then again, if it were a bad patch, the patched CCE 2.67 would spoil the whole video stream and WOULD NOT select certain frames to spoil them to reveal its pirate protecton scheme or whatever..
Why shouldn't it?

If you don't have a legit CCE versions, your comparisons are worthless. Sorry to say so!

[r6d2]

Originally posted by Yuri
with the logo and without the logo, patched and not patched, I don't care a hook about that all
Well, the forum really cares and when registering you agreed to abide by the rules.
and the result is that 2.50 IS better than 2.67 in any respect, which is TECHNICALLY proved.
You know, Yuri, you were conducting with great ability a very interesting experiment which got lots of attention. But the thing you have best proved so far is that you care little about respecting a mod and the forum rules you agreed to.

It is a pity, since you were showing good wood.
Any discussion about pirated software, mystical pirate-protection behavior of this software is irrelevant.
It may be irrelevant to you, but the fact is that you recognize to have been using illegally obtained functional copies of a software, AKA crack (not just "trials"), and that is not allowed in this forum.
to share my technical experience but not ethical or moral one
In spite of that, you have clearly showed what ethics mean to you.

[Matthew]

Okay Yuri, I'll glady admit that I'm wrong =) I'm confused as to why CBR would also exhibit the same problem though, but no matter.

Back on topic - with all this patching stuff fortunately confined to the dustbin - perhaps the test should be expanded to some other sources...anime/cartoon is not typical I suppose. And also multipass...how many people use 1-pass anyway.

[r6d2]

Originally posted by Yuri
Dear Moderator, then again, what are we discussing here? Moral principles or technical stuff?
Technical, certainly. Principles are taken for granted.
I thought I could bring some useful information to people but I'm afraid I had been wrong!
Actually, I think you were doing just fine.
have the same logic as this[...]
Mmm... Your logic, however, is interesting indeed, it's like this:

I agreed to behave according to community standards, which I read and understood. Then just because I'm so cool, I don't give a hook about that. Somebody reminds me about my inconsinstency, and I pick up my ball and go home.

"Those are my principles. If you don't like them, I have others."
-- Groucho Marx
I retire [...]morality is taking its turn now, that's not for me.
We know, we know. You are right this time. You are not bringing useful information to people.