Inquiring on HC Encoder built-in matrices

[2COOL]

Anywhere I can get a full description on what these matrices can do for me. All I see is numbers. It's not explained in HC's PDF.

I've gathered some info from perusing the forum and the replies in this thread.

HC
NOTCH
QLB
BACH1: Preferred bitrate range [ 1000 - 1800 ]
JAWOR1CD: Preferred bitrate range [ 1500 - 2500 ]
HVSGOOD
HVSBETTER
HVSBEST
AUTOQ2 (aka AVAMAT7): Preferred bitrate range [ 1800 - 3500 ]

Custom matrices
AVAMAT6: Preferred bitrate range [ 3500 - 9500 ]


I assume all matrices mentioned in PDF are working properly with no bugs?

[Amnon82]

AutoQ2-Matrix is one of mine. It is a matrix you can use for the most motion pictures. If You have CCE 2.70 the encoder calculates it during the encode (AQM). It also works with HC realy good. Thats why Hank added it to his program. He liked it.

It is gread in the bitrates around 1800 to 3500. I also used it for lowbitrates in TMPGEnc: 500 - 800.

In higher bitrates every matrix will do almost the same picture quality (Thats what I think).

To know exactly what the matrix will bring for You, download LigHs matrix-test-tool.

CQME - Editor (http://forum.doom9.org/showthread.php?t=73666)

Also there You can find my matrices AutoQ1, AutoQ2 and AVAMAT6.

... I don't remember the compare tool of ligh. Ask him. I search the forum, but I didn't find it today.

[dragongodz]

The so called QLB Matrix IS the KVCD Notch Matrix.
2COOL - i really expected better of you than that. just reading the thread you referenced you can see that is NOT the case. the notch matrix just was accidentily not replaced in 1 QuEnc release by me. so certain people jumped up and down that i was somehow trying to claim/steal their precious notch matrix even though i corrected the mistake the very next morning(so it was something like less than 12 hours between releases).

for more information on the QLB matrix you can read this thread.
http://forum.doom9.org/showthread.php?t=91821

[2COOL]

@Amnon82

Thank you for your input.

@dragongodz

I apologize for causing you more grief. :( I was merely speed searching through forum and came up with whatever. I didn't fully read the other posts. I was looking for statements. My reference is removed in your cause.

My sincere apologies and here...have some bubblegum on me. thanks for the link.

[feedback]

@2COOL
This info. I got from the Rebuilder Matrix Editor by Rockas on the
Jawor's1CD matrix...recommended for between 1500-2500.

"This quantization matrix is intended as a replacement for the H.263 and HVS - Good Picture matrices. It should give good results when encoding natural movies at low bitrates. Preliminary tests have shown a 8% gain of compressibility compared to the HVS - Good Picture matrix andabout 4% compared to H.263. Note: First non-intra value should be 16... meaning that this Matrix may be incompatible with some Players".

Regards,:)

I use AVMAT6 which is DVD compliant, a lot with DVD-RB Pro. using the HC encoder. I haven't used Jawor's 1CD.

[Amnon82]

@Feedback: Thanks for using my Matrix AutoQ1 (AVAMAT6).
@2Cool: AutoQ2 (AVAMAT7) is the matrix for higher compression. AutoQ1 (AVAMAT6) I did for MEncoder. This encoder didn't handle matrices over 62.

AutoQ1 (AVAMAT6)
8,16,19,22,26,27,29,34
16,16,22,24,27,29,34,35
19,22,26,27,29,34,35,38
22,22,26,27,29,34,35,40
22,26,27,29,32,35,40,48
26,27,29,32,35,40,48,50
26,27,29,35,40,48,50,60
27,29,35,40,48,50,60,62

16,20,24,28,32,36,40,44
20,24,28,32,36,40,44,48
24,28,32,36,40,44,48,52
28,32,36,40,44,48,52,56
32,36,40,44,48,52,56,58
36,40,44,48,52,56,58,60
40,44,48,52,56,58,60,62
44,48,52,56,58,60,62,62


AutoQ2 (AVAMAT7)
8,16,19,22,26,28,32,38
16,16,22,24,28,32,38,44
19,22,26,28,32,38,44,48
22,22,26,32,38,44,48,54
22,26,32,38,44,48,54,64
26,32,38,44,48,54,64,74
32,38,44,48,54,64,74,84
38,44,48,54,64,74,84,94

16,20,24,28,36,42,46,52
20,24,28,36,42,46,52,58
24,28,36,42,46,52,58,62
28,36,42,46,52,58,62,68
36,42,46,52,58,62,68,78
42,46,52,58,62,68,78,88
46,52,58,62,68,78,88,99
52,58,62,68,78,88,99,99

[2COOL]

Thanks for the extra information Amnon82! :D Can I bother you to divulge a little more information on your matrices? if not for me, then for other newbies coming unto the scene.

What I mainly wanted to inquire on was the average bitrate range a specific matrix would be optimal in and if possible, for what type of movie it'll likely be intended for. e.g. action, anime, etc.

You mentioned earlier that AutoQ2 is great within 1800 to 3500 but what about AutoQ1? I'm still trying my best to interpret the numbers but it's hard at first glance.

I apologize for not searching thoroughly. I jump between an average of three DVD backup forums daily, reading and helping, and it almost seems like another full-time job. I've already overloaded my life a few times by adding some programming projects, beta-testing other people's tools, and writing guides at the same time too.

[Boulder]

Jawor's matrix was designed for XviD so it might not be that good for MPEG2.

[Amnon82]

@2Cool:

AutoQ2Matrix is great within 1800 to 3500, AVAMAT6 will do 3500 to 9500 best, cos it dosn't compress so much. AVAMAT6 is DVD compliant. AutoQ2Matrix should be also.

I'm traveling much via trains. During a travel AVAMAT6 came into my mind and I wrote it down. After I came home I tried it with MEncoder and got good results. Later I coded a predictiontool for CCE called AutoQ, so I renamed AVAMAT6 to AutoQ1.

AVAMAT6 was a popular name. I only renamed it cos I changed my nic from Avalon to Amnon82 (some bad stuf from the kvcd.forum).

AutoQ2Matrix (or AVAMAT7 if You like) was designed for higher compression. I began with VCDs and TMPGEnc. AVAMAT1 to AVAMAT5 I experimented with matrices in TMPGEnc.

CCE handles numbers higher than 62 so I adapted AVAMAT6 for CCE with higher compression but same structure.

I looked on many matrices and this came out during traintravels and thinking about numbers ;)

I'll do some encodes with AVAMAT6 and AUTOQ2Matrix and upload the screenshots to my homepage www.autoq2.fx.to.

I'll let You know

[Asmodeus]

Jawor's matrix was designed for XviD so it might not be that good for MPEG2.

But it is. I use it all the time, and there isn't any incompatibile issues.

[2COOL]

Thanks Amnon82 ;)

I'm also glad you clarified the matrix names as I would've been more confused thinking that there are four matrices instead of two out there.

AutoQ1Matrix = AVAMAT6
AutoQ2Matrix = AVAMAT7

[Amnon82]

K, I won't confuse You 2Cool. My matrices are known as AVAMAT6 and AutoQ2Matrix. I edited the post above to get rid of the confusion ;)

[feedback]

I like the AVAMAT6 name instead of the AutoQ1Matrix. (That name is more well known I believe).

I was using the BDVD Matrix a lot and I still like it for more compression (it has a somewhat smoothing effect) but the AVAMAT6 is better with detail, it is like the new standard mpeg matrix but with better compression and...well it just looks better to me.
Actually someone else in the DVD Rebuilder forum recommended it to me a while back.

Amnon82, I had no idea that AutoQ1Matrix = AVAMAT6, but I have them both and just checked and of course you are correct they are the same except in name.

Maybe you should stay with the AVAMAT name as people will recognize that the same person that did 6, did 7, etc.
Just my opinion.

Regards,:)

P.S. I may give the AVAMAT 7 a go, I just copied it to notepad then saved it as AVAMAT 7, then imported it into the Rebuilder Matrix Editor located Here. (http://dvd-rb.dvd2go.org/modules.php?name=Downloads&d_op=getit&lid=103) .
The Rebuilder Matrix Editor will allow you to convert from .cqm, .xcm or text files BTW.
NEURON2 and his
QuantDump (released under GPL - source code available
at:http://neuron2.net/dgmpgdec/quantdump10.zip) - This
utility is the responsible for the Matrix "extraction"
from the original Video Stream is a neat feature also.

[Amnon82]

I did a 22SecEncode with a scene of speed. I uploaded it already to megaupload, but forgot to save the link. So I must upload it again. I'll tell You when I did it, again ;)

AutoQ2 Encodes (http://www.megaupload.com/?d=38TPJEE7)
AutoQ2 Encodes - Missing Part (http://www.megaupload.com/?d=26HCIQZ7)

Screenshots can be find at www.autoq2.fx.to (http://amnon82.cybton.com/AVAMAT/AVAMATS1.html)

[2COOL]

Screenshots can be find at www.autoq2.fx.to (http://amnon82.cybton.com/AVAMAT/AVAMATS1.html)I like your screenshots but it is possible for you to zoom in on particular spots to emphasize on existing artifacts while comparing it with the original?

Here's an example (http://forum.digital-digest.com/showthread.php?t=54724).

[Amnon82]

Hmm. I uploaded all my encodes to megaupload. I think it is the best to see it with Your own eyes. 100 MB is not so much. ... or do a encode with AVAMAT6 and 7.

[ljpp]

The fact that matrices for HC (and others) are so poorly documented has been bugging me for some time. And when you ask the experts for recommendations, all one gets is stupid aswers like "there is no best" or "depends on the matrial" etc. useless crap.

Some sort of chart with recommendations to target bitrates is highly needed for the less experienced, and the ones that actually do have a life and don't have time to perform n^x test encodes before performing the actual encoding task.

Based on the crumbs I have found here:

>3500kpbs MPEG
<3500kbps QLB
>1800kbps AutoQ2

I would gladly publish such a "FAQ chart" of matrices, if I would just receive sufficient background info from the community.

:goodpost:

[feedback]

The Rebuilder Matrix Editor in the link that I posted above has a lot of info. on Matrices.
For example:

Angle BestLow:
The Angel-matrices were used initially for minimizing interlocks at standard SVCD bitrates. For this purpose they completely quantize the high frequencies out of the picture. The advantage: The compressing on macroblock level functions better, which saves bitrate. Disadvantage: The picture loses sharpness, in addition Rainbow-effects or blocks in soft backgrounds can occur. This matrices are also suitable for the Extras were we decrease the bitrate. Best Low approx. 1900 - 2500 kbps

BDVD:
This matrix was created by the Brazilian BDVD Team and they claim that this is the matrix that will give the best results. But I must warn you that the first "Intra" value doesn't fit on the mpeg standard (16 should be 8). This can cause problems on some players. From the tests I made, I can say that it worths the risk, the results are pretty good.

Kika Comic Low CG:
Bit rate saving without end. Very well in order to remove/reduce noise, but produces some quantization errors. They become noticeable as visible blocks in monochrome faces, however only at the monitor, not on TV. But for that, as well as no blocking with scenes in motion. Good for for digital sources (CG) on a lower bitrate.

Regards,:)

[Amnon82]

@feedback: And what do You think about AVAMAT7 (AutoQ2Matrix)?

[feedback]

@feedback: And what do You think about AVAMAT7 (AutoQ2Matrix)?
I have not had the chance to use it yet...

I will let you know when I do.

Regards,:)

[CirTap]

I would gladly publish such a "FAQ chart" of matrices, if I would just receive sufficient background info from the community. and I'd be the first to bookmark it! :D seriously.

I feel your pain as this "big secret of recommended bitrates" (we're not allowed to say the B-word here) has been bugging me since I learned DVDs are made of compressed video.
The "experts" probably see instantly what a matrix is for, know their pros and cons and write each om them down by heart, but not everyone doing DVD has a binary brain :) and as you perfectly pointed out: ... and the ones that actually do have a life and don't have time to perform n^x test encodes...

Thanks to LigH's editor (thumbs up!) I was finally able to get a sort-of-a-grasp of what those damn numbers could "possibly" do.

In summary: :goodpost: @ 2COOL for starting this!

@ljipp: hope the thread evolves and serves as a good source for your chart :)

Have fun,
CirTap

since I honestly give a sh** about tweaking every single bit out of my video to squeez 'em even on a 5" floppy disc, I use MPEG with "full rates" in almost all encodes: DVDs aren't that expensive anymore and I can easily spend a second disc for the extras & specials if I think they're worth keeping

[ljpp]

Unfortunately, as always, this discussion never picks up. I dont understand why this is not published as a FAQ or something.

[Carpo]

Based on the crumbs I have found here:

>3500kpbs MPEG
<3500kbps QLB
>1800kbps AutoQ2

:goodpost:

i tend to use

main = HVS good
<3mb = mpeg standard
<2mb = QLB
<1mb = bach1

but after reading about these other 2 matrices i might give them a go instead :)

is ConvertToYUY2() required to be ticked when you use HC or is that just a CCE requirement

[Carpo]

have done a few test encodes using avamat7 matrix (aka AutoQ2-Matrix) using 7 passes on cce and 2 passes on HC (i know more than 2 is not really worth it and it looks an unfair test - but all will be clear)

i havent included the times as hc will be faster due to less passes, also have not used ConvertToYUY2() when using HC as after a bit of reading have been told its only required for CCE which i did use it on

(forgot to add NaNs dll and avisynth 2.56 used)

Test 1

1min 30sec clip

CCE SP 2.70 Trial - no filters -7 passes - AutoQ2-Matrix (AVAMAT7)
HC 0.15 - no filters- 2 passes - AutoQ2-Matrix (AVAMAT7)
(15a errors out on here - and i like not to use 0.16test_release as its still being tested)

both were great looking - although i feel HC looked a little crisper than CCE encode

so 1-0 to HC - so far

Test 2

UnDot and Deen were used on this test as the film i am going to do has a rather high compression ratio (54.2%) although it not being grainy or noisy that much there was a bit there, and i have been advised to use Deen on high compression films

1min 30sec clip

CCE SP 2.70 Tial - 7 passes - AutoQ2-Matrix (AVAMAT7) -UnDot().Deen()
HC 0.15 - 2 passes - AutoQ2-Matrix (AVAMAT7) - UnDot().Deen()


once again HC looks spot on - colours and overall picture look good and doesnt look as soft as normal when i have used Deen

CCE looks just as good as HC does - although i did notice that HC encode did tend to be a bit larger than CCE - thus possibly causing encodes to be bigger than 4,35GB - which i have had only twice with HC

I would have to say (being a CCE man) that HC does give just as good a result if not better than CCE SP - so im going to guess that it is also better than the basic version

If you want to save time then i would say use HC - if you dont mind how long your encodes take to use CCE or AutoQmatEnc (By Sapstar) as that is also a good encoder although a bit slower than most - although quality is good

Finally - i guess i will have to add that i am concidering going over to HC :p

:helpful:

[Amnon82]

Great test for my AVAMAT7 (aka AutoQ2-Matrix). Nice to read You like it. HC is great. Hank asked for my matrix as he saw it ...

[Carpo]

have edited my post above, renamed the matrix to the right name ;) - and thought it was time i tried to give back to a place where i have taken so much from :devil:

also hope hank releases 0.16 soon i used the test release earlier today and its looking good

[Amnon82]

There was no need Carpo. Hank will rename AutoQ2-Matrix to AVAMAT7 in the next release, cos everybody likes the well known name AVAMAT. My old matrix was called AVAMAT6. If you're reading my earlier posts in this thread you know more on this.

... there is also no need to rename it back to AVAMAT7, only if you want to :D

[Carpo]

it was late i was tired :sly:

[DK]

and i like not to use 0.16test_release as its still being tested

i have used 016 tr2 many times now and so far i havent come across a single problem at all

and i should think this was meant to be just the other way around, wasnt it?
If you want to save time then i would say use HC - if you dont mind how long your encodes take to use CCE

cce is faster but i like hc better especially in lower bitrates

[dragongodz]

i have used 016 tr2 many times now and so far i havent come across a single problem at all
there is still a problem with the ### ###. oh cant say anything about that yet can i. ;)

this was meant to be just the other way around, wasnt it?
DK - you must have missed that Carpo used 7 passes for CCE. lets all chant "overkill", "overkill","overkill". :D

[Pasqui]

@dragongodz
Do you mean that it is not safe to use HC0.16test_release2 ? Is this issue minor or major ? Thanks for your advice,
Pasqui.

[Carpo]

t
DK - you must have missed that Carpo used 7 passes for CCE. lets all chant "overkill", "overkill","overkill". :D

that was done to point out something - hc will give the same result as cce - so it showed that HC will deliver what CCE SP can give you at a fraction of the time and passes (and cost :) )

i have used 016 tr2 many times now and so far i havent come across a single problem at all

and i should think this was meant to be just the other way around, wasnt it?


cce is faster but i like hc better especially in lower bitrates

I never use beta or test releases of software as they may contain bugs which could possibly effect the outcome of the test - so i always use stable releases (although cce cant always be called stable) :D

[dragongodz]

Do you mean that it is not safe to use HC0.16test_release2 ? Is this issue minor or major ?
its a test release so meant for testing. if you check out the proper thread for it you will see there are not to many problems. the things holding up a 0.16 proper release are not related to that. so yes it should be pretty safe to use but there still may be some hidden bugs that people havent come across. of course this is true of full releases aswell so that shouldnt worry you too much. :)

that was done to point out something - hc will give the same result as cce - so it showed that HC will deliver what CCE SP can give you at a fraction of the time and passes (and cost :) )
its only a fraction of the time because you used a ridiculously high number of passes for CCE though. if CCE actually NEEDED that many passes then that would be one thing, however it doesnt. a VAF + 2 pass encode should give you a practically same quality encode from CCE. there have been tests done in the past to show that. so if you want to compare speed then you should test CCE with a sensible number of passes.

ummm and this is all way off topic for a thread about matrices. :eek:

[DK]

sorry for being OT once again, but YES you're right:

DK - you must have missed that Carpo used 7 passes for CCE. lets all chant "overkill", "overkill","overkill"

i did NOT read that and it explains alot :p

[ljpp]

Amnon,

Hank will rename AutoQ2-Matrix to AVAMAT7 in the next release, cos everybody likes the well known name AVAMAT. My old matrix was called AVAMAT6.

Ever considered more descriptive names for these? I mean, for the Joe Average (sorry Joe), these avamats and autoQs don't mean sh*t.

Still very few recommendations or suggestions for matrices to be used at various typical bitrate levels. Are the people unwilling or unable to post their suggestions so that we could gather them together and reach some kind of concensus?

[CirTap]

*bump*