QMatOp for DVD-RB : QMat Optimization

[raddygast]

Right you are. It's just that QMatOp, being what it is, has sparked all this discussion of "maximum" quality. Otherwise we'd all be using DVD Shrink or Nero Recode or some other transcoder. So it just seemed natural to start discussing whether CCE really is the best option, or if it's worth switching to HC. Using QMatOp already adds so much time to the process that I don't mind using a slower encoder.

I am really glad I found DVD-RB and all the other awesome tools I've used so far. Now the only killer feature I'm waiting for is for some way to run multiple projects with QMatOp. I don't even mind if a given film takes 8 hours total to process (prepare/qmatop/encode/rebuild) -- as long as it can do that unattended.

[SAPSTAR]

Originally posted by raddygast
....I'm waiting for is for some way to run multiple projects with QMatOp. I don't even mind if a given film takes 8 hours total to process (prepare/qmatop/encode/rebuild) -- as long as it can do that unattended.
Be patient it may come :)...The next version will add a new feature of auto GOP (useful for CCE, useless for HC), and most probably a GUI. Then the version after will most probably allow you to run it in batch !

[hutch1711]

SAPSTAR, am I right to assume it really doesn't matter what matrix is used in the DVD-RB prepare stage as QMatOp will do it's own thing and workout the best matrix, regardless of which matrix was initially used?
Also, is there any way QMatOp might adapted to AVI material, or for that matter any other source material not using DVD-RB in the prepare stage, being coverted to DVD?

[Carpo]

yes - is the answer to the first question as i asked excatly the same thing :D

interesting point on your second question tho :P

[SAPSTAR]

Originally posted by hutch1711
SAPSTAR, am I right to assume it really doesn't matter what matrix is used in the DVD-RB prepare stage as QMatOp will do it's own thing and workout the best matrix, regardless of which matrix was initially used?
Also, is there any way QMatOp might adapted to AVI material, or for that matter any other source material not using DVD-RB in the prepare stage, being coverted to DVD?
For your first question, I say NO, explanation :
If no matrix is specified, then it will do everything by itself.
If a matrix is specified, then it will be taken in to account in the final average. Depending on the framesplitperiod you chose, its role will change....smaller is the framesplitperiod, smaller is the importance of this matrix.
Conclusion: You can specify a matrix in DVD-RB, if you want to give a small hint to QMatOp, otherwise you don't care.

About the AVI, well it could be adapted (I suppose Divx or Xvid)....but I may have to develop either a separate exe, either a special mode.

@carpo : you didn't exactly ask that, you asked me if it was working without choosing a matrix in DVD-RB. Anyway, with a framesplitperiod < 15, it doesn't really matter that you specify any matrix. :)

[hutch1711]

Thanks for the quick reply and I'm looking forward to your next version of QMatOp with auto GOP. This program is evolving so quickly you dare not blink :)

[Carpo]

Originally posted by SAPSTAR
For your first question, I say NO, explanation :
If no matrix is specified, then it will do everything by itself.
If a matrix is specified, then it will be taken in to account in the final average. Depending on the framesplitperiod you chose, its role will change....smaller is the framesplitperiod, smaller is the importance of this matrix.
Conclusion: You can specify a matrix in DVD-RB, if you want to give a small hint to QMatOp, otherwise you don't care.

About the AVI, well it could be adapted (I suppose Divx or Xvid)....but I may have to develop either a separate exe, either a special mode.

@carpo : you didn't exactly ask that, you asked me if it was working without choosing a matrix in DVD-RB. Anyway, with a framesplitperiod < 15, it doesn't really matter that you specify any matrix. :)

well i asked about vbrbias in dvd-rb and what qmatop would do with it - thought same might have applies to matrices :o

also on a iff subject - it seems you cant use ms java - installed vs 6 earlier - which installs ms java - to test i remove suns version and told xp to use ms version of java and qmatop didnt like it :D

[Carpo]

think may have found a poss bug

a friend of mine ask if i could do a back up of one of the films his parents sent him from japan seeing as his pc is broken (dont ask what its called as i couldnt tell as its in japanese :P), no prob i thought - ran it thru dvd-rb and it looks to have made 65 segmants.

now when qmatop hits the 53rd segmant it just freezes sticks on

Pass 1/1 frame 12/20, 3.21fps, ETA: 00:00:02.496, size: 0.00MB - repeats this line continuously

yes i have pre processed which is why i say poss bug (and also could be issue my end - although this is the only film i have had issues with) - thing is tho if it does this many segmants on main movie - if i leave the rest in how many more will it make :/

i will try in the morning on un prorcessed film and let you know what happens

dvd-rb 0.90apro QMatOp 0.12b cce 2.67

[SAPSTAR]

@carpo: thank you for info about ms java...(not really surprised in fact :) ) - let me know about your pb. Are you able to play the AVS file from the concerned segment (with any video player... I mean)???

[Carpo]

yip i can play the avs files ok - did a radom check on a few of them - im checking now using unprocessed source and if worse comes to worst i will re-install as i have install a few things that i wouldnt norm install (vs6 sdk headers) and poss wondering if it could be the left overs of that causing it

will have a bit of a muck about with it and let you know

also any eta on v0.13 :cool:

well tried it on un processed and it still sticks at

Pass 1/1 frame 12/15, 0.00fps, ETA: 00:00:00.000, size: 0.00MB

on segmant V01005100001052.AVS - its only 1 sec long tho - dunno if that will have any impact on the issue, have also noticed that avs2avi.exe hits around 50% cpu useage and just sits there but runs fine up till segmant 52- also i will try a diff film see if it really is my pc or not

will see what you have to say b4 i re-install just in case ;)


right ran final fantasy thru dvd-rb and then ran qmatop no issues baring in mind there was only 47 segmants for it to look at - so its either the amount of segmants thats an issue or the fact that its only a second long

[SAPSTAR]

Originally posted by Carpo
...right ran final fantasy thru dvd-rb and then ran qmatop no issues baring in mind there was only 47 segmants for it to look at - so its either the amount of segmants thats an issue or the fact that its only a second long
Amount of segments...nope ! I already handled this case in one of the very first version. About the one second long, I also managed that. I really don't see what it can be, because avs2avi is used in rather simple way, execution of an AVS script to sample the original one....there must be something wrong in the installation...could you try on another PC ?

[EDIT] Stupid question : do you have enough free space where is located QMatOp ? Did you reboot ? Did you try to reinstall Sun Java ?

[Carpo]

Originally posted by SAPSTAR
Amount of segments...nope ! I already handled this case in one of the very first version. About the one second long, I also managed that. I really don't see what it can be, because avs2avi is used in rather simple way, execution of an AVS script to sample the original one....there must be something wrong in the installation...could you try on another PC ?

[EDIT] Stupid question : do you have enough free space where is located QMatOp ? Did you reboot ? Did you try to reinstall Sun Java ?

have about 150GB left on the drive where QMatOp sits

yes i uninstall Sun and MS Java and re-installed sun

im gussing its a source issue as i have done FF fine 4 times this morn

[hutch1711]

This may not be an error, but I thought I should report it anyway. I backed up a movie using HC, changing the QMatOp RangeLimit to 255. When I compared the matrix values computed in the Rebuilder ECL file, with those extracted from the M2V files, by a program called MatrixExtractor 1.10, any value over 127 in the intra-frame side is shown as follows 128=0 129=1 130=2 131=3 132=4 etc.

[hank315]

Just checked Matrix Extractor for values > 127.
For non-intra correct values are displayed, for intra it screws up if values > 127.

[hutch1711]

So intra values >127 are being coded correctly by QMatOp, and the bug is in Matrix Extractor?

[SAPSTAR]

Originally posted by hutch1711
So intra values >127 are being coded correctly by QMatOp, and the bug is in Matrix Extractor?
...are coded correctly by HC !!!! QMatOp only passes matrices to HC...it's not the encoder !

[hutch1711]

Originally posted by SAPSTAR
...are coded correctly by HC !!!! QMatOp only passes matrices to HC...it's not the encoder !
I knew that! My dummy :)

[Carpo]

ne news on a new ver ?

[SAPSTAR]

Originally posted by Carpo
ne news on a new ver ?
Well...the GUI is finished. But FishMan0919 found a bug on National Treasure....and I'm waiting for a test from him...The Auto-GOP feature is not yet fully tuned. It should be a question of one or two days before the next release.

[Carpo]

perfect timing that is.......should be a few days before my hdds defrag and i can start to backing up dvds again :D

[SAPSTAR]

First a special thanks to FishMan0919 who tested a lot of alpha version before finding a pb with National Treasure.

Changes in 0.13b :
- Added an Auto GOP feature - Still being tested and to be tweaked.
- Created a GUI for starting the whole process - Still to be improved
- Added a progress bar
- Improved the way smallest segments are managed. Should be faster now.

You start QMatOPGUI, then go !!! Everything will start automatically once the Prepare phase is finished. Don't panic if you can't see any window...QMatOp is in the background waiting for DVD-RB.

The AutoGOP function is still being tested...I will most probably tweak it in a near future.

[Carpo]

testingout the new autogop on a pr0n film - if it screws it up (no poun) no loss :P

seems this version is rather fast took about half the time to do the same film than when i used 0.12b :)

[Axlemar]

Originally posted by Carpo
testingout the new autogop on a pr0n film - if it screws it up (no poun) no loss :P

lol at least your honest. Anyway, I will be sure to use the new version in the next day or two. Thanks for the work on it.

[hutch1711]

Sapstar, I know you mentioned in this forum that the new auto GOP feature would not work with HC, but I notice it's not noted in the version history :)

[SAPSTAR]

Originally posted by hutch1711
Sapstar, I know you mentioned in this forum that the new auto GOP feature would not work with HC, but I notice it's not noted in the version history :)
You're right...too be more precise, HC has an auto GOP function which is already included and most probably better than the one from QMatOp...I think that HC is optimizing the GOP structure for every GOP, in my case I can only optimize for each segment...
So if you activated the QMatOP's AutoGOP, it will be ignored in HC but it won't bring any issue.

[artoor]

I have one question :) What will happen if I'll push "stop/pause" button in DVD-RB when it will be working along with QMatOp to resume encoding process later (or next day)?? Should I open "QMatOPGUI" before resuming again??

Sorry for (probably) weird question but I'm layman in this case. Nevertheless I'd like to try this - as many people say - great software and get to know more about QMatOp :)

Regards!

[SAPSTAR]

Originally posted by artoor
I have one question :) What will happen if I'll push "stop/pause" button in DVD-RB when it will be working along with QMatOp to resume encoding process later (or next day)?? Should I open "QMatOPGUI" before resuming again??

Sorry for (probably) weird question but I'm layman in this case. Nevertheless I'd like to try this - as many people say - great software and get to know more about QMatOp :)

Regards!
If you stop the encoding, just restart it later...but don't restart QMatOp again( The ECL file is already adjusted) ! The only issue you may have is with the AutoRebuild, not sure if it would work....

[artoor]

If you stop the encoding, just restart it later...but don't restart QMatOp again( The ECL file is already adjusted) !
Uhmmm... the ECL file is already adjusted so I can continue encoding process even after restarting my PC (but without opening QMatOp again as You've said). Have I understood You right?

The only issue you may have is with the AutoRebuild, not sure if it would work....
Let's assume worse option (sorry ;)). If it wouldn't work. Will be there a chance to backing up my DVD until the end?? (by hand instead of auto like just is in DVD-RB 3click mode) Can You tell me?? (I've never used QMatOp so forgive me my concern)

Thanks for Your answer :)

[SAPSTAR]

Originally posted by artoor
...Have I understood You right?...
yep !

.. If it wouldn't work. Will be there a chance to backing up my DVD until the end?? (by hand instead of auto like just is in DVD-RB 3click mode) Can You tell me?? (I've never used QMatOp so forgive me my concern)

You just have to manually press Rebuild in DVD-RB as if QMatOp was never used.

[artoor]

One more time - Thank You!! :)

[raddygast]

Well, it pains me to have to post this, but I've been using QMatOP for at least a week now, have encoded several films, including all six discs of the Lord of the Rings extended editions.

Unfortunately, I was not satisfied with some blockiness that I noticed in a few scenes. So I re-encoded one of the discs without QMatOp and viewed one or two scenes as a frame of reference for comparison (Two Towers Disc 2, at around 1:39:47, as well as some other scenes with Elvish subtitles).

The result was that using CCE encoder default looked far better than the QMAT optimized output. The QMat version was much blockier. This was confirmed by using the "Digital Zoom" feature of PowerDVD at 4x. The QMat version looks horrible zoomed in, but the Encoder Default version looks much much better.

The same was true for Elvish subtitles (in the video source, not a subtitle stream). There is a definite "halo" of blockiness or noise around the subtitles. It is most minimal in the original source (not re-encoded). The encoder default version is a bit more blocky, but acceptable. The QMAT version is the most blocky and messed up. Again, I also checked with zoom at 4x.

I understand the zoom algorithm could be faulty, but it was just confirming what I saw with my own eyes before zooming, so I tend to trust it.

Now I have no argument with QMatOP as a program -- I thought it was a fantastic idea and am glad for SAPSTAR's effort. Especially nice was the auto gop and auto quality prec/auto vbr bias. But if the results are going to look like this, I may as well just save my time.

Now I realize that maybe there are parts of the film that greatly benefit from QMatOp, where the Qmat-opped output will look better than the encoder default output. But I checked some more "action" type sequences and the default version always either looked better to my eyes, or the same.

Is this perhaps a case of my having a preference for "smoothness" vs. sharpness? Anyway, if anyone wants to check, the main test I used besides subtitles was the scene at 1:39:47, where Merry and Pippin are in the flooded spoils of Isengard, and the camera is slowly zooming in on them (before it goes to a full close-up on Pippin). Merry's face, in the distance, looks really blocky, and much better in the encoder default version.

Mind you I preprocessed the source and removed a bunch of audio tracks, and used MenuShrink to get the menus down to almost nothing. But the bitrate was quite ample, something like 4155 avg. I was very disappointed with how the QMat version looked.

NOTE: Both encodes were done with the EXACT same pre-processed source tree, so that can be eliminated as a variable. TargetSectors was set to the same thing as well in both RB projects.

Is this a case of actual practice disproving the theory (the theory being that QMatOp uses perceptual error algorithms to optimize the output in the best way possible, whereas the CCE default matrix is a compromise)?

I'm open to discussion. I'm not trying to trash this, just wanted to be a heads-up because at first I thought with all these rave reviews that there was NO way QMatOP didn't improve quality.

[Carpo]

do you have a lcd monitor ?

i have and sometimes films do look blocky - yet when i play them on standalone player and 42" tv they look fine

few ppl i know say that lcds screens are not as forgiving as norm tvs

just my 2 pennys worth :)

[raddygast]

No, I watch DVD's on a 19" CRT. It's true that on a non-HD TV, from far away, the blockiness is probably not visible, but the issue is that it's there to begin with. Just because I may not be able to *see* a loss of quality from a certain viewing configuration doesn't mean it's irrelevant. It's sort of the principle of the whole business that has my panties in a twist.

[onesoul]

I'm sorry but I have to agree with raddygast. Maybe the problem is dctune itself.

[archaeo]

Well I must say that I am glad to be reading some feedback in this thread like raddygast's that discuss actual quality results obtained with QmatOP. Somebody doing some actual side by side comparisons. I haven't had the time to beta test the program myself, but have been looking through this thread for some feedback that indicates the results are worth the extra processing time. In theory it sounds like an awesome program, and I'm grateful for SAPSTAR's work on this, but up to this point there hasn't been much dialogue on this issue.

[raddygast]

Yeah, I mainly posted my bit in the hopes that someone who swears by QMatOp (Fishman maybe) can give some examples of its performance. Maybe if we all had the same frame of reference, say a common movie with a particularly difficult scene to encode.

I just went by the assumption that QMatOp knows what it's doing. I seriously doubt it has to do with algorithm bugginess, but maybe there's something flawed in the approach (not the algorithm). Comparing perceptual error for still frames seemed at first like a great idea, but maybe that's not the best way? I have no idea what other technique would be better though.

In theory, wouldn't it be possible to get a series of say 30 frames that differed (perceptually) minimally from the source, but when played together over the course of a second might give the impression of "blockiness" or "fuzzy noise" ?

Seriously, I have no clue about the theory behind matrices and so forth, so I don't know how much of this makes sense. All I can remark on is the quality that I noted. I don't want to rain on this parade AT ALL -- just want to be pointed in the right direction. Maybe those scenes bugged me, but there may be a bunch of other scenes in the film where QMATOP really shined.

Still, for noww I won't be using it until further investigation.

[Axlemar]

I didn't see anyone state what encoder version they were using or what setting they used. I thought the quality was good with 1%/5 seconds and CCE 2.70. Try more aggressive settings and see if it is still worse.

[borgraf]

First of all, kudos to SAPSTAR for working on this very cool tool :)

Originally posted by Axlemar
1%/5 seconds and CCE 2.70. I used exactly that on two test sources and compared them to runs with QLB, which won both times.

The problem with my QMatOp runs can be decribed as a lack of detail. One was Alexander R2, ~3060 bps with partly stripped menu. One scene in the beginning had a harbor looking like a still, another scene about 1h 22m in has a wide shot with little movement of Alexander in a room with all his commanders; when he turns his head it's "jerky" and looks pretty bad. The QLB run was fine on both these examples and more. I'll be testing some more in the near future. Maybe the problem is only with lower bitrates?

I agree with Carpo that LCD screens are not as forgiving. So a film will look "better" on the TV, but the LCD is also great for comparing runs and finding "flaws" :)

Keep up the good work coders and testers!

[SAPSTAR]

@raddygast : What were your settings in QMatOp ???? QMatOp was highly tested and optimized with 2% / 10 fps. I'm not sure if by going under the 10 fps the formulas are acting well.

You said you tested the material on a 19" CRT ? It's a REALLY small screen, I wonder how you could notice anything ?!?! Myself I'm using a 120" screen with an Infocus projector...and I didn't have the kind of issue you talking about ?!?! Pay attention not to activate the ANIMATION mode. About the AutoGOP, I would say...avoid it for now, it's not really stable.

@borgraf : You're saying that QLB was better than CCE SP 2.70 (with or without QMatOP) ?!?! [EDIT] About low bitrates, if you are using HC + QMatOp, you'll have VERY good results at very low bitrates, even with Half-D1.

@all : I think you misunderstand what QMatOp really does, it optimizes the Quantization Matrices in the REBUILDER.ECL, to really compare what it does you have to keep the same encoder in each test !!!! I mean, process a movie with CCE SP alone, then QMatOP + CCE SP, otherwise you're comparing apples with pears. By the way, you can use QMatOP with any encoder in DVD-RB : HC, QLB, CCE SP, it doesn't matter.....That's why I really don't understand your test : (CCE SP 2.70 + QMatOP) vs QLB ?!?!
Try (QLB + QMatOp) vs QLB or (CCE SP + QMatOP) vs CCE SP or (HC + QMatOp) vs HC........

[Fishman0919]

@ All

If you are view your movies mostly on your computer... QMatOp might not be for you and you might have better viewing with out it. Some of the movies I have tested didn't look as good as I hoped on my computer, but viewing them on my 65" Pioneer Elite HDTV... 9 out of 10 movies looked as good if not better (to me). I had a professional technician come to my home and calibrate my tv to as close to NTSC standard as possible and I am view my movie in an optimal environnement(dark with minimal light).

Some people like a smooth more flat picture... Qmatop might not be for you scene it will give you a sharp, detailed picture... maybe to sharp sometimes. Lets remember its a beta and every movie might not look great. I have tested a lot of movies for SAPSTAR and a few looked pretty bad.. but after some tweaking he got them to look better.

If QMatOp doesn't work for you, then it doesn't work for you, you don't have to use it... it's an add-on to a great program to help people who want a better quality encoding... but if it doesn't work it doesn't work... it adds time to the encoding process that you don't need and you can enjoy your movie sooner.

If you like, please comment on ways to help improve the results and quality by posting what movies and settings and/or matrices you used. Please don't make insults like some person who I won't mention.

Like I said it's a program to help you with the quality of you encoding but if it doesn't work for you, sorry.... but if it does...Enjoy.

[raddygast]

2%/12 was what I used. Anyway, it's whatever setting is the default in QMATOP.ini. I didn't change it (though I thought about doing so). Also, I used the same encoder consistently, on the same sources. And Animation was not selected.

19" is a small screen, but I should have clarified it was a computer monitor, not a TV. As such I sit much closer to it. As Fishman has now allowed, qmatop results may not look so great on a computer screen, maybe even worse than default CCE (as is my experience). He contends that it can look better than default on an actual TV screen viewed from a proper distance. I can't argue with this because I have no means to try it.

I can groove with the option than qmatop results may be "sharper", i.e. perhaps a little too sharp sometimes. Certainly up close to a computer monitor, I prefer smoothness. But perhaps on one of those insanely expensive screens like you have, I'd notice the difference and nod in favour of QMATOP.

Anyway, like I said I'm not here to bash. I love the idea of stuff being figured out for me automatically. But I guess for my current purposes and setup, default is better (for now).

Finally -- what do you guys think about DVD-RB's "CCEAQM=1" option you can insert in rebuilder.ini? Apparently it tells CCE to adjust qmats automatically and dynamically within each cell. What principle does it work on?

[Fishman0919]

@ raddygast,
Sorry my point was not pointed at you as a source of insults and sorry if you thought so...

CCEAQM=1 or Adaptive Q-matrix Switching is, in short, a way that CCE SP can automatically adjusts and creates a new matrix for each GOP if the lowest quantization scale cannot be raise to the minimum bitrate under a single quantization matrix. I find that having this turned on help on some movie with qmatop, adjusting the matrix if qmatop has not got it just right. It will help with any martix, CCE Standard, MPEG Standard, QLB... some say that they don't like it but it can only help... when the encoder needs help on a hard to encode GOP, Adaptive Q-matrix Switching adjust the matrix for better quality and only then.

I really don't see why it should be turn off... but that my opinion. :D

[Axlemar]

But if you turn Adaptive Q-matrix Switching on with the default matrix and turn the animation on in qmatop wouldn't all the matrices calculated by qmatop be overwritten? Also, wouldn't Adaptive Q-matrix Switching kind of defeat some of the purpose of using qmatop since it is supposed to be kind of similar to it? If not, then the only reason people shouldn't use Adaptive Q-matrix Switching is because some players will give odd results or skip like crazy.

[Carpo]

Originally posted by raddygast
2%/12 was what I used. Anyway, it's whatever setting is the default in QMATOP.ini. I didn't change it (though I thought about doing so). Also, I used the same encoder consistently, on the same sources. And Animation was not selected.




did you use that for all films ? ntsc and pal - seeing as i dont know what format you have im only guessing here but if you used 2/12 on ntsc that might be a reason for your poor results

as has been posted before norm its 2/12 for pal and 2/15 for ntsc


if you have used the right settings above ignore my post :D

[raddygast]

They were NTSC, but SAPSTAR previously posted that half the framerate is a good compromise between speed and quality. He said that using 2/12 was foolproof in that it was half of PAL and interlaced TV sources, and it was perhaps a bit slow with NTSC 29.97 fps films but certainly not poor quality -- if anything, better quality than running 2/15.

But maybe I misinterpreted him.

[Carpo]

what i always do is check the vobs using dgmpgdec130 if it says pal 25fps - i use 2/12 if it says 29.97 - which it does 9/10 i use 2/15

and i never seem to have any probs - although i have just experienced my first oversized disc but dunno if thats cus of qmatop or HC encoder which i have been using for last few films - so have to get shrink out to squash it down :/

[borgraf]

Wow.. tough crowd in here :)

1) By QLB I mean the Quenc Lower Bitrate matrix supplied with DVD-RB, and that I used it with CCE 2.70. So I was indeed comparing apples to apples.

2) I never use my monitor to view entire films

I thought I was fairly clear, but I guess i'll have to be more specific in the future. Atleast in this thread :)

[Carpo]

where in the rebuilder ini would CCEAQM=1 go ?

options or cceoptions ?

[archaeo]

<options>

[SAPSTAR]

@raddygast: About the results on a computer monitor, I'm not surprised as this kind screen is not managing the video as a reg. screen. I know by experience, that some well known matrices give artifacts on Computer screens, but those artifacts disappear on a a regular screen.....
Everything is based on the perceptual view of pictures, but I suspect that the perceptual view is not the same on a computer screen. So QMatOP must not be for you if you're watching DVDs on your computer.
(For example, with interlaced movies it's horrible to watch them on computer screen, you watch them on a TV.....and taaadaa perfect.)

@borgraf: I didn't understand that you used the QLB matrix on the same encoder. I suggest you to try : 1%/10fps instead of 5fps....then let me know your results. I'm really surprised because up to now I had really good results with almost all movies I did.

@all : About the AutoQMat in CCE SP, it can cause pb in some DVD players because the matrix is not changed at each cell, but at each GOP !!! And I think, that's it's only adjusting it a little bit so if the passed matrix is already good, it will be able to adjust it more accuratly.

@all : Something else, I'm ready to try to tweak QMatOp, if you can provide more info on the pb....QMatOp settings, movie, symptom, ....

[EDIT] I may for example, add parameters to change your screen config. Some parameters in the whole algo are depending on "physical" parameters....(dpi, distance from the screen...)