QMatOp for DVD-RB : QMat Optimization [Archive] - Page 6

Fishman0919

15th May 2005, 15:01

@ Axlemar

No, only when CCE SP can't raise to the minimum bitrate under a single quantization matrix... that may happen for 50 GOP's in a row, it may happen for the whole movie or it may not happen at all. The closer the matrix is to being correct for the GOP being encoded the less CCE SP will have to adjust the matrix with Adaptive Q-matrix Switching if at all.

I have a couple of DVD Player including the 2nd DVD Player Pioneer made (DV-05), one of the pickyest DVD Player there is... no problems with qmatop.

hutch1711

15th May 2005, 17:07

Sapstar, I just want to say I think this is a great tool!
I've had excellent results backing up region 1 material using 2% tolerance and a framesplitperiod of 15. I've had some issues with picture quality from contouring noise in flat wide areas (looks like a speading puddle of water), since QualityPrecAuto and VbrBiasAuto were added, but not on all sources. Switching these features off in the .ini file, and using only the tolerance and framesplitperiod features, usually corrects the problem. One of the nicest aspects of this tool is the ability to switch on and off certain properties and adjust others.

raddygast

15th May 2005, 17:13

that puddle of water effect sounds like what I was getting on the within-video-source subtitles. The original source had them, faintly, but QmatOp made them much more noticeable.

Axlemar

19th May 2005, 04:32

(edited) nevermind

borgraf

20th May 2005, 17:09

Originally posted by SAPSTAR
I suggest you to try : 1%/10fps instead of 5fps....then let me know your results.During the week i've been running 9 different encodes of a PAL project, 3h12m at 2950 bps "typical". It gives me a chance to test and compare both QMatOp and HC to my usual routine with CCE or ProCoder. I'll do some more comparing over the weekend, but what i've seen so far the combo of CCE 2.70 4pass+QMATOP (with QualityPrecAuto=0, VbrBiasAuto=0 and 1%/10fps) finished in last place. Second from the bottom is the same setup except for QualityPrecAuto=1, VbrBiasAuto=1. I did not use GOPAuto or Animation.

Compared to the other encodes, the QMatOp runs have more "artifacts" throughout and one segment in particular stands out as looking fairly bad. It's encoded at vbr_brate_avg=2103 and QMatOp produced this matrix:

qmat=
8 11 11 11 17 64 123 123
10 11 11 13 17 42 123 124
11 12 18 24 37 123 124 124
13 14 28 123 123 124 124 125
16 26 123 123 124 124 125 125
105 123 123 124 124 125 125 126
123 123 124 124 125 125 126 126
123 124 124 125 125 126 126 127
21 120 121 121 122 122 123 123
120 121 121 122 122 123 123 124
121 121 122 122 123 123 124 124
121 122 122 123 123 124 124 125
122 122 123 123 124 124 125 125
122 123 123 124 124 125 125 126
123 123 124 124 125 125 126 126
123 124 124 125 125 126 126 127

The first run used DVD-RB defaults and the second run QMatOp produced these values:

quality_prec=21
vbr_bias=10

I'll compare some more over the weekend, maybe I can come up with something more useful :)

SAPSTAR

21st May 2005, 00:11

@borgraf
What is the movie ???
Did you try HC + QMatOP with a 255 limit + AutoVBR??
On low bitrates that's the combination I found more efficient....

Oh..something else REALLY important on what kind of screen did you compare the results ?
I ask that because some matrices can show artifacts on a computer screen and are perfect for TV screens......I may have to introduce a parameter for that, it's depending on the screen DPI.

borgraf

22nd May 2005, 10:48

Originally posted by SAPSTAR
@borgraf
What is the movie ???It's a project of the Extended Edition of Lord of the Rings - Fellowship of the Ring, 2 discs joined to 1.

Did you try HC + QMatOP with a 255 limit + AutoVBR??
On low bitrates that's the combination I found more efficient....What's AutoVBR? I have a run with HC at BEST and QMatOp with RangeLimit=255. Here's the matrix for the same segment as in my above post:

qmat=
8 11 11 12 17 65 180 181
10 11 11 13 17 42 180 181
11 12 18 24 37 180 181 182
13 14 29 167 181 182 183 185
15 28 166 181 183 184 186 188
105 179 180 182 184 187 191 194
130 180 181 183 186 191 196 202
181 181 182 185 188 194 202 211
21 136 136 136 137 138 139 141
136 136 136 136 137 138 139 142
136 136 137 137 138 139 141 143
136 136 137 139 141 142 145 147
137 137 138 141 144 147 150 154
138 138 139 142 147 152 158 164
139 139 141 145 150 158 167 176
141 142 143 147 154 164 176 192

This also have the "artifacts" that only the QMatOp processed tests have. I would describe it as funny edges and things shifting in and out of focus. My clean runs with CCE 4 pass, HC at BEST and ProCoder 2 at Mastering Quality shows none of these.

Oh..something else REALLY important on what kind of screen did you compare the results ?
I ask that because some matrices can show artifacts on a computer screen and are perfect for TV screens......I may have to introduce a parameter for that, it's depending on the screen DPI. I'm using both monitors: 17" TFT and 19" CRT, but mostly a high-end 32" CRT TV. I can see the same things on the monitors as on the TV, of course the TV is a bit more forgiving. In my experience that is always the case, and QMatOp makes no exception.

SAPSTAR

22nd May 2005, 13:01

@borgraf : By looking at the inter matrix,....I can tell that QMat had to compress at the maximum this part....it indicates a very low bitrate for highspeed scenes. How long is the movie ???

@all : Would you be interested if I introduce 2 new parameters : Screen Size + Screen Quality level ???? It may solve certain kind of visual issues for some people......

Carpo

22nd May 2005, 14:19

i would say yes but would also say to fix autogop first :)

jdobbs

22nd May 2005, 14:57

Originally posted by SAPSTAR
@borgraf : By looking at the inter matrix,....I can tell that QMat had to compress at the maximum this part....it indicates a very low bitrate for highspeed scenes. How long is the movie ???

@all : Would you be interested if I introduce 2 new parameters : Screen Size + Screen Quality level ???? It may solve certain kind of visual issues for some people...... Making the output to a DVD tailored to a specific screen type? That doesn't sound like a very good idea...

borgraf

22nd May 2005, 17:18

Originally posted by SAPSTAR
@borgraf : By looking at the inter matrix,....I can tell that QMat had to compress at the maximum this part....it indicates a very low bitrate for highspeed scenes. How long is the movie ??? The answers to both your questions about bitrate and film length are in my previous posts. And my encodes with CCE, HC and ProCoder without QMatOp used the same bitrate and came out just fine.
Originally posted by jdobbs
Making the output to a DVD tailored to a specific screen type? That doesn't sound like a very good idea...
amen to that :)

Carpo

22nd May 2005, 18:24

thats why they are (hopefully) selectable options ;)

use them at your own peril :D

Carpo

25th May 2005, 10:29

range limit atm is set to 127 (default) but according to the mpeg standard how far above that could you actually go ?

Fishman0919

25th May 2005, 10:35

Originally posted by Carpo
range limit atm is set to 127 (default) but according to the mpeg standard how far above that could you actually go ?

Well 127 is the default because CCE SP only supports that.... MPEG Standard is 255... HC Encoder will do that.... not sure about QuEnc

Carpo

25th May 2005, 10:54

even new cce ? if so stupid cce ;D

Fishman0919

25th May 2005, 11:02

Originally posted by Carpo
even new cce ? if so stupid cce ;D

From the CCE SP 2.70 user guide

Create quantization matrices
On Quantize matrices setting window, you will see two matrices, one
for Intra block and another for Non-intra block. You can set these
matrices separately. To modify a matrix, type the value directly on
the matrix.

Allowable range of the value is : 1 to 127 (3).
(1, 1) element in intra block must be 8 under MPEG standard.
And the buttons below a matrix will help you to modify the value.
• button copies (i, j) element and put them in (j, i) when i < j.
• button copies (i, j) element and put them in (j, i) when i > j.
• Transpose button interchanges rows and columns in a matrix.
It means the interchange of (i, j) and (j, i) elements.

(3) Cinema Craft Encoder SP limits the value to 127 in spite of MPEG standard
255.

yes, very stupid cce :angry:

Carpo

25th May 2005, 11:11

bummer :angry:

guess i give HC a go and set it as 255 :)

SAPSTAR

25th May 2005, 16:40

@all : I'm not releasing anything new because I'm currently working on an encoder. And because QMatOp seems to be rather stable, I'll add new functionalities once I'll find time to do it ! :)

alfixdvd

26th May 2005, 20:13

Please help me

When I run Qmatop I get java.Io.IoException error=2

Fishman0919

26th May 2005, 20:18

Originally posted by alfixdvd
Please help me

Help you with what???

alfixdvd

26th May 2005, 20:20

Sorry

I edited the post

SAPSTAR

26th May 2005, 20:26

Originally posted by alfixdvd
Sorry

I edited the post
Please paste the whole message......

Hostile_Monk

27th May 2005, 00:58

Thanks for all the hard work Sapstar, can't wait to try out the encoder!

alfixdvd

27th May 2005, 07:50

Originally posted by SAPSTAR
Please paste the whole message......

The problem was a corrupt cmd.exe

I replaced it with the cmd.exe of another PC and all's right.

Thanks

alfixdvd

27th May 2005, 17:13

Why when QmatOP starts after Prepare phase of Dvd Rebuider I get the message:

Can't find the file c:\DvdToos\QmatOp\frames\Intra*.ppm

Qmatop continue working normaly.

SAPSTAR

27th May 2005, 17:58

Originally posted by alfixdvd
Why when QmatOP starts after Prepare phase of Dvd Rebuider I get the message:

Can't find the file c:\DvdToos\QmatOp\frames\Intra*.ppm

Qmatop continue working normaly.
It just makes some cleaning before starting....but most of the time useless. The message is just a warning not an error....

alfixdvd

27th May 2005, 18:38

Thanks Sapstar

SAPSTAR

27th May 2005, 19:38

@all : alfixdvd discovered a bug with the latest version of DVD-RB (0.93.2).
A correction is available on my website......

:thanks: alfixdvd !

unbx

28th May 2005, 09:59

Hi, i recently used this program and i found with one of my dvds, that the bitrate flew everywhere, from 4.5Mbs to 11.0 like every few seconds. This may not be qmatop but i found this odd and i found that on one of my dvd players, the sound cut out where the bitrate was at its highest.

Just wondering if this is abnormal.

Lo_RaT_PeNaT

28th May 2005, 15:54

SAPSTAR, as I have written in another post, I use the version 93.2 PRO, of the rebuilder and I/you had not observed anything strange, but when reading that written by Alfixdvd, let us stay observing the behavior of the DVDRBInterface.exe, and so much the one that comes included in the version that you can discharge in your you paginates, as the new one that you have put, I observe that Prepares when finishing the phase and to begin the qmatop, the DVDInterface.exe, disappeared of my toolbar, as finishing your function, to my I/you seemed normal, until I have used another version of the Rebuilder again, in short the For 0.92, in this case, the icon doesn't even disappear in the phase Encode, for what I must suppose that something happens, he/she looks at this that I comment you, because you really happens this way, and you even happens me with the new version that you have put in your you paginates of the Interface, you even disappears during the first phase. A greeting waits you understand what I am commenting you.

alfixdvd

28th May 2005, 16:17

if with the new DvdInterface and DvdRebuilder 0.93.2, Qmatop's windows closes as quick as finishes Prepare phase of DvdRebuilder. Please check the next:

workdir in qmatop.ini points to the correct working directory of DvdRebuilder
MPEG2DEC in rebuilder.ini points to DgDecode.dll in DvdRebuilder's directory.

Lo_RaT_PeNaT

28th May 2005, 16:24

Rebuilder ini

[Options]
CCETargetSectors=2268965
HCTargetSectors=2261000
CCE=3
LoadPlugin=1
Completed=106
EncoderMinimized=1
HalfExtras=0
idct7Opt=1
DECOMBFlag=1
Mode=5
OneClick=1
HalfD1=00
Convert_16_9=00
DisableInterlace=10
Reminder=80
SkinVersion=5
HC_Quality=3
HC_OPV=1
DCPrec=3
MovieOnly=0
ISO_Output=0
iDCT=7
[CCEOptions]
VBR_bias=20
Quality_prec=12
eclPasses=6
OPV=0
[Paths]
CCENEW=C:\Archivos de programa\Custom Technology\CCE SP Trial Version\cctspt.exe
Source=D:\PELICULAS\EL AVIADOR REMAKE\
Working=D:\PELICULAS\AVIADOR\
MPEG2DEC=C:\Archivos de programa\DVD-RB PRO\DGDecode.dll
DECOMB=C:\Archivos de programa\DVD-RB PRO\Filters\Decomb521.dll
REJIG=C:\Archivos de programa\DVD-RB PRO\Encoders\ReJig\ReJig.exe
Output=D:\PELICULAS\AVIADOR\
HC=C:\Archivos de programa\DVD-RB PRO\Encoders\HC Encoder\HCbatch.EXE
QuEnc=C:\Archivos de programa\DVD-RB PRO\Encoders\QuEnc\QuEnc.exe
ProCoder=C:\Archivos de programa\DVD-RB PRO\Encoders\EclPro\EclPro.exe
[Audio]
Selected=11
[Subpictures]
Selected=111

As you can observe in the rebuilder ini, I have on position the dgdecode.dll, in the place where you indicate, and I assure you that anything but to begin the qmatop, the DVDRBInterface.exe, disappears in the version 0.93.2 For of the Rebuilder, I didn't seize in the version 0.92 For

Qmatop.Ini

workdir=D:\PELICULAS\aviador\D2VAVS
tolerance=2
framesplitperiod=12
AutoEncode=1
AutoRebuild=1
Animation=0
RangeLimit=255
QualityPrecAuto=1
CCESP26X=0
VbrBiasAuto=1
GOPAuto=1

hutch1711

28th May 2005, 17:12

SAPSTAR

30th May 2005, 18:29

I have been using QMAPOp on a number of test DVDs using different encoders. Each time I have to run QMAPOp after the DVD-RB preparation stage to optimize the matrices. Since (and I assume this is correct) the matrix values are the same, I would like to save the matrix values produced the first time around on subsequent encodes, using different encoders. Any ideas? Or maybe it just can't be done. Thanks.
It could work if DVD-RB was not refusing it :) I mean that once a REBUILDER.ECL file has been created...if you change the kind of encoder, DVD-RB is refusing to start with the current REBUILDER.ECL it asks for a new prepare phase...so you lose all the matrices :( Maybe I should backup the matrices in separate file, to be able to reuse them....sounds possible...

[EDIT] If you are really courageous, you can make a backup of the REBUILDER.ECL, then paste the matrices in the new REBUILDER.ECL :devil:

hutch1711

30th May 2005, 22:35

Sapstar :thanks:

curious d

6th June 2005, 21:09

Sorry about the troubleshooting post in the other thread. I hope this is the appropiate thread to troubleshoot. I configured the QMatOPGUI accordingly. I unrar'd parts 1 and 2 into the same QMatOP folder in the DVD-RB folder. I have unclicked the one click option. I restarted my computer and it still isn't an automated function for me. DVD-RB will prepare the first VOB file then stop. QMatOP does not automatically start for me. I have to click on the QMatOP icon to get it to run. How does one get it to run for the entire movie and not just for the first VOB file? Sorry for the ignorance. I'm certain this seems all so very basic to everyone else, but I'm getting the feeling that I'm missing something very simple. Thanks.

hutch1711

6th June 2005, 22:11

Curious D, the problems you are having are not likely caused by QMatOp. QMatOp is run after DVD-RB completes the preparation stage, and if I hear you correctly DVD-RB processes the first VOB file and then quits. QMatOp, which you will start manually after the DVD-RB preparation stage, hasn't even begun to work it's magic yet. Your problem is probably in your DVD-RB setup, and it's been suggested in other posts that you use Rockas Installer for DVD Rebuilder to ensure it is configured correctly.

curious d

6th June 2005, 23:41

Thanks for response. I should correct myself in saying that the preparation stops after the first VTS. That leaves the extras and the menu behind (with the movie only selection NOT selected). When I got the Pro version, there were two versions available. The installer version and the zip version. I used the installer version. I don't know if it was there is a Pro version with Rockas installation unless the installer version was the Rockas version. Certainly, the installer version did not come with the bundle that the freeware version came with. Does Pro come with that bundle?

Incidently, I am able to make a back up of my discs without QMatOp if that is of any help to my query.

Thanks again.

hutch1711

7th June 2005, 02:04

If you downloaded the pro version .93 exe it includes the rocka files. I'm confused by your statement of being able to backups without QMatOP. QMatOp isn't required by DVD-RB to produce backups. It's a helper program to optimize the matrix values using the data produced by DVD-RB in the preparation stage. Are you saying that you previously backed up the disc that is currently giving you trouble with DVD-RB?

curious d

7th June 2005, 05:56

Acutally, my question really is about using QMatOp. My understanding to this program is that it should used when DVD-RB is in the three click mode. I thought that when the prepare phase is done, then QMatOp should start automatically. But it doesn't for me and so I start it up. After QMatOp finishes its task, then I press encode. I have read that after encoding, some users get QMatOp to run again at this stage before rebuilding. Is this necessary to get optimal results?

My initial concerns about getting only the first VTS file did not pan out as when it finished, the whole movie was in place. So now my question only pertains to its use as stated above. Thanks.

By the way, I also was wondering what the difference between part 1 and part 2 of the downloads for QMatOP? Thanks again.

BadServo

7th June 2005, 08:21

Forgive this question as it is most likely one of ignorance. After reading through the relevent posts, I've got the recommended QMatOp parameters for NTSC/PAL and Interlaced/Non-Interlaced seared into my brain. I'm also aware of what parameters are ignored by non-CCE encoders. However, I'm curious about the absolute best parameters to use, assuming processing time was of no consequence.

I should say up front that after a week of tests churning out comparison samples on 3 PCs, I'm a die-hard HC Encoder user. Am I correct in making the following assumptions based on what I've read?

-When using HC, use 255 as Range Limit
-Set Frame Split and Tolerance to 1
-Diable the AutoGOP, as HC's is presumed superior

From what I can gather, these settings would give the longest processing time, but also the best possible quality and would be suitable for all source material.

Any opinions or corrections would be very graciously accepted.

Djuby

7th June 2005, 08:51

Hi everybody,

Can someone recommend matrix setup of DVD-RB for HIGH/LOW/TYPICAL Bitrates: 3,645/120/2,894 or I should just use decoder default?

Djuby

SAPSTAR

7th June 2005, 18:11

...
-When using HC, use 255 as Range Limit
-Set Frame Split and Tolerance to 1
-Diable the AutoGOP, as HC's is presumed superior
....

If you're ready to wait for a long time : BINGO ! :) Let me know the results I'm interested in it......

SAPSTAR

7th June 2005, 18:13

Hi everybody,

Can someone recommend matrix setup of DVD-RB for HIGH/LOW/TYPICAL Bitrates: 3,645/120/2,894 or I should just use decoder default?

Djuby
Hmmmm depends on the movie...but as a general advice I would try :
HVS Good Picture for Main Movie
AVAMAT6 for >3000
QuEnc for lower bit rates
and Bach1 for extras.....

Djuby

7th June 2005, 21:41

This is a series disk (Star Trek Enterprise). I only use CCE for encoding (IMO the best encoder out there). My question is more like, out of the preconfigured matrices that come with DVD-RB, which one should I use for the above mentioned bitrates?

Djuby

BadServo

7th June 2005, 22:22

Well, a while ago I finished up my 22 hour-long QMatOp session. Here are my results.

Source Length: 3 Hours
Source Format: NTSC Interlaced
Encoder: HCBatch 0.14

I have to admit that this test really surprised me. I am suspecting that I chose a poor source material for the test, as it is animation, and older animation at that, making ti harder to judge quality.

I ran this encode 3 times, each time using HC Encoder on the "Best Profile". The first encode was done without QMatOp at all, using that encoder defaults for Matrix settings.

The second run was encoded using QMatOp with what I'll call teh "standard" settigns for this type of source (2,15,127) with CCE specific options disabled. Quality was quite good. However, for the most part I noticed no significant quality change between using these settings and using the HC defaults. Again I blame the source. The on exception was during a scene with rapid motion of white-blue light on a pitch black background. There was some visible artifacts at the intersections of color luminosity.

The third run used the insane settings of (1,1,255). Not unexpectedly, the the resutl was very much the same. In fact Identical to the second attempt. Although the artifacting from encode 2 did seem to be slightly diminished.

Based on my 3-person comparison test, I feel that this test would definately have been better served by some clean CGI source material with sharp contrasts. When time permits, I'll rerun it. I feel confident in saying that there's no real reason anyone should need to use such high-settings for QMatOp, but to each his own. ^_^

I also want to note, that on every previous test comparing encodes with and without QMatOP (using both CCE and HC) QMatOp has always produced a superior image on lengthy source material. This is the first descrepency in that "norm."

hutch1711

7th June 2005, 22:44

I thought that when the prepare phase is done, then QMatOp should start automatically.
No, QMatOp must be started manually.
After QMatOp finishes its task, then I press encode.
If you check off the AutoEncode and AutoRebuild boxes in the QMatOpGUI the process will be done automatically.
I have read that after encoding, some users get QMatOp to run again at this stage before rebuilding. Is this necessary to get optimal results?
I have never tried this so I'll let Sapstar answer this one.
I also was wondering what the difference between part 1 and part 2 of the downloads for QMatOP?
Just a guess, but I believe there are 2 parts because it has something to do with being a rar file.