TMPGEnc: File Size Test with Differents Adjusts

[jorel]

i did this tests using TMPGEnc versions 2.58.44.152 and 2.521.58.169
and the results are exactly the same in quality and sizes :

tmpgenc adjusts for kvcd 352x240 mpeg1 lion king 3

first test and settings adjusts:

video tab:
CQ 90
B picture spoilage 0 (zero)

GOP structure tab:
detect scene change

Quantize matrix tab:
use floating point DCT

final size 593Mb

second test changing only this settings adjusts, the remainder are the same:

CQ 97
B picture spoilage 20

GOP structure tab:
detect scene change

Quantize matrix tab:
use floating point DCT
no motion search for still picture...

final size 572Mb


more CQ, better quality , less size

believe, in the whole cartoon the image is better in the second encode !
i did 2 times each test with this 2 tmpgenc versions...no changes!


comments are welcome
thanks.

[jorel]

i forgot important detail:

when i got 593Mb with CQ 90, in the script was
Asharp (1,2)
in the first encode

then i reencode using CQ 97 and
Asharp (1,3)
in the second encode with differents settings
and was waiting more final size.....that seems logical in first sign.

asharp (1,3) will encrease the quality and size,.....this i was waiting.

for this reason i posted that the image is better and with less size!

[Prodater64]

Hi: And really are you who are asking for batteries?

I don't know to much about this, but I think your results are related to b picture spoilage, what permits a percentage of corrupted b pictures.
(Excuse me I don't know if you already know that). But in any case, this is the way that MoleVCD or MVCD how you prefer work around less file sizes. I don't know how are its effects on final quality, one thing is what you see, and another one that it is measurable. (But I'm sure in your case is almost the same, my dear friend daylong kvcd maker)

[jorel]

Quote:

Originally Posted by Prodater64

Hi: And really are you who are asking for batteries?

I don't know to much about this, but I think your results are related to b picture spoilage, what permits a percentage of corrupted b pictures.
(Excuse me I don't know if you already know that). But in any case, this is the way that MoleVCD or MVCD how you prefer work around less file sizes. I don't know how are its effects on final quality, one thing is what you see, and another one that it is measurable. (But I'm sure in your case is almost the same, my dear friend daylong kvcd maker)

don't have doubts,
i don't knew about b picture spoilage, and less from MoleVCD (i never see or use it)


effects on final quality?
like i posted the second encode with less final size is better in pc and tv


"....one thing is what you see..."
my target is the best result that i can see !
and yours ?

[Prodater64]

Take a look
http://www.kvcd.net/forum/viewtopic....t=molevcd+mvcd
http://www.kvcd.net/forum/viewtopic....t=molevcd+mvcd

And jorel, what are happend to you? Whats a mad?
Are you fine, not healthy problems like arterioesclerosis?
See this http://www.kvcd.net/forum/viewtopic....t=molevcd+mvcd

[jorel]

Quote:

Originally Posted by Prodater64

Take a look
http://www.kvcd.net/forum/viewtopic....t=molevcd+mvcd
http://www.kvcd.net/forum/viewtopic....t=molevcd+mvcd

And jorel, what are happend to you? Whats a mad?
Are you fine, not healthy problems like arterioesclerosis?
See this http://www.kvcd.net/forum/viewtopic....t=molevcd+mvcd

arterioesclerosis?
maybe....
but i wrote about mole:
"i never see or use it"
means:
i never see any encode and never see the results and never use it before and never will..
and......oh no, the arterioesclerosis is back...i forgot the rest!




still kiding like in my last post but... it's true

[Prodater64]

Quote:

Originally Posted by jorel

but i wrote about mole:
"i never see or use it"
means:
i never see any encode and never see the results and never use it before and never will..
and......oh no, the arterioesclerosis is back...i forgot the rest!

You are worse than I think.
If you thought "i never see any encode and never see the results and never use it before" but only wrote ""i never see or use it", not only arterioesclerosis but a kind of "dislalia" too.

[jorel]

oh sorry, i think that my horrible english was clear for you!
i don't use translations like you used here:
http://www.kvcd.net/forum/viewtopic.php?t=10022

my english is horrible and......i'm proud of it !
...
what means " dislalia " ?


i'm worse than you think?
well, depend of your referencial for "better" and "worse" ....
your referencial is truty ?
what's " worse " for you?
the bad that you are feeling or the bad that i'm not feeling ?
i have hope doc?....or " better " is a confinement?!?!?



psychology is temporal.....today the doc think this, tomorrow that....
but the results of the tests with tmpgenc don't.

"better" is back to the topic, right ?

[Dialhot]

Hi guys.
I didn't read all the threads, just your tests with b spoilage Jorel.

I thought you were aware of that big thread "CQ vs CQ_VBR" that contains 39 pages of messages !
In this you can find a part were it is explained that the "b spoilage" setting BLOWS the CQ curve towards the bottom.

In others words CQ90/B=0 is a better value than CQ97/B=20.

In fact I never did tests with such high CQ but for CQ=70 and B=20 I already saw that I have ht same result with B=0 and CQ=65 !

[jorel]

Quote:

Originally Posted by Dialhot

Hi guys.
I didn't read all the threads, just your tests with b spoilage Jorel.

I thought you were aware of that big thread "CQ vs CQ_VBR" that contains 39 pages of messages !
In this you can find a part were it is explained that the "b spoilage" setting BLOWS the CQ curve towards the bottom.

In others words CQ90/B=0 is a better value than CQ97/B=20.

In fact I never did tests with such high CQ but for CQ=70 and B=20 I already saw that I have ht same result with B=0 and CQ=65 !

yes, i remember that big and important thread and need to read it all agian.
is a good reference for "no lazy" newbys.....they will learn a lot!

only comments...
dvd2svcd 113b3beta7 and older versions, use B picture spoilage to 20 too ,
..in my hd with backups, have some d2s.ini used in old encodes .

don't know if you read the second post with more details...
not only the CQs used but Asharp too...i was waiting more size!

you know, all recomendations are welcome and
if you have some more.....great!

thanks !

[Dialhot]

You do not have a such big difference into the 2 scripts. From memory, the second parameter of the ashap is the radius; The first one is the strength !

In other words asharp(1,2) or asharp(1,3) is slightly the same thing but asharp(2,2) produces a big difference !

[jorel]

Quote:

Originally Posted by Dialhot

You do not have a such big difference into the 2 scripts. From memory, the second parameter of the ashap is the radius; The first one is the strength !

In other words asharp(1,2) or asharp(1,3) is slightly the same thing but asharp(2,2) produces a big difference !

yes Phil,
asharp(2,2) produces a big difference and is not usable for this case.
the image got heavy contrat for 352x240 and maybe in other resolutions too.

i did little tests(trim with 1000 frames) for that 4 situations:

with B=20
asharp(1,2) 6,23Mb --->same quality as B=0
asharp(1,3) 6,27Mb --->same quality as B=0

with B=0
asharp(1,2) 6,33Mb --->same quality as B=20
asharp(1,3) 6,37MB --->same quality as B=20

the differences from B=0 and B=20 are imperceptible in the images,
but "too big" for this short samples !
remember, it's only tests (that i do all the time),
with B=20 and asharp(1,3) got 6,27MB

seems the same quality like in my 2 full encodes
that i posted in the first post!

your recommendations:
trash B=20 and change to B=0
B=20 can be like a "trap" (snare) for some scenes or sources!?!?

[incredible]

a) Asharp(1,2) does smooth plain parts of the image (based on threshold which is here low=2) more than Asharp(1,3). Thats why asharp is NOT like unsharp mask in Photoshop, means accord. to asharp() - the lower the threshold gets, the more spatial softened gets the image at surfaces = risky in case of hair details etc.! Do your test and see the difference of Asharp(1,4) and Asharp(1,1) (thats the main reason why asharp(1,2) gives you less filesize but therefore more details killed )

b) MVCD 2.6 (where the templates are done by roc and myself ) do have a B picture spoilage because on the other hand less frames in a gop are used, means 15 at PAL and 18 at NTSC which does compensate this and on the other hand an non-intra matrix which is especially designed to work in combination with B pircture spoilage. ..... just to clearify that as these 2.6 Templates are not moles work anymore but roc's and mine as Mole never have been seen there anymore and these templates by this can be seen as (Martin)vcd and Inc's*lol* as Martin is roc's real name. (and before a big asking chainreaction starts: Quality is almost the same)

Don't use Bpicture spoilage in case of Notch/24 GOP ... as KVCD Templates are set optimal as they are.

[Dialhot]

Quote:

Originally Posted by incredible

and before a big asking chainreaction starts: Quality is almost the same

The same as the first horrible MVCD ?

[jorel]

thanks ink !

"Don't use Bpicture spoilage in case of Notch/24 GOP ... as KVCD Templates are set optimal as they are."
ok ink, but i'm using GOP 15 for this cartoon, i still don't have to use B=20 ?

"MVCD 2.6 (where the templates are done by roc and myself )...
....as these 2.6 Templates are not moles work anymore.."
right....i really don't knew that, nothing against but i never used it before! "

"...Martin is roc's real name."
where is Martin?
why he don't post more in english forum?

ps:
roc's and scharfis_brain's avatars are really cool !


<edited>
hey ink,
Phil is back "baby" !

[incredible]

Quote:

Originally Posted by Dialhot

Quote:

The same as the first horrible MVCD ?

From the donkey network?? :P
(no not like the "first" MVCD ... as Im far away from this and as Im knowing that community only since the new templates where made - I meant the same Quality compared to KVCD before a big comparison starts which would be nonsense )

Quote:

hey ink,
Phil is back "baby"

What?? He was lost??

Roc I know very well and he is a very moderate a friendly guy and Sharfis_Brain also .. very helpful and a "fieldbased acting and thinking" person with knowledge on this as hell. He takes much of his time to analyse samples from other peoples at doom9.org when they got problems with strange interlacing and fieldblendings

[jorel]

yes,i know about roc and scharfis_brain works and tests.
really cool!
call roc to post in english forum too!

i always like the roc's avatar and the new sharfis_brian avatar version is amazing too.

yeah, Phil is back and
mcvd !

(kidding)

but....and the answer to my doubt ink ?!?!?

you wrote:
"Don't use Bpicture spoilage in case of Notch/24 GOP ... as KVCD Templates are set optimal as they are."
but i'm stup and ask for more details:
ok ink, but i'm using GOP 15 for this cartoon, i still don't have to use B=20 ?


thanks in advance again!

ps:
all the time i have to post "thanks" for you,Kwag, Phil, and for all.
maybe is a good idea put an "unremovible thanks" in my signature, don't?

then my keyboard remains working for loooong time!

[incredible]

According to BpictureSpoilage .....:
Only your Eyes can tell you the truth "for you". So IF you like more the image seen in a whole, then choose Bpicturespoilage and on the other hand gain some from the better CQ.
As every setting in mpeg compression is a compromise you just have to choose

Its a logic, .. the less Bframes you got within a whole gop count, the less risk you get of visible quality loss of BframeSpoilage. But as said above your eyes will decide.


BTW: The Spirit from KVCD changed very much as now Quality is the main purpose than the compressibilty (Yep Kwag I know you intended the best compromise but you will understand when reading the next few words). As many People of doom9.org/de do think KVCD is just an out of standard way to compress and not keeping quality --- they should take a look at the mencoder threads in here ... I do NOT know any other place where new mpeg1/2 methods and things are tested down to the bones that much!

[jorel]

very clear ink !


like i wrote, B=20 can be like a "trap" !

thanks !