Next changes to optimal script.
 

Thanks to Jellygoose for the suggestion on trying out unfilter once again, instead of using asharp's undocumented negative values for blur.

Here's the latest experiment:

ScriptClip("nf = YDifferenceToNext()"+chr(13)+ "nf > 2.5 ? unfilter( -(fmin(round((nf/0.5)), 100)), -(fmin(round((nf/0.5)), 100)) ) : \
TemporalSoften(2,7,7,3,2) ")

It seems to do a better job than asharp, specially on the scene changes. Give it a try, just to make sure it works ok. It might also have a "positive :?: " impact, with higher accuracy on file prediction.

Thanks for the suggestion Jellygoose ;)

-kwag

Ok, will asharp still be the sharpener in the script :?:
or do we use unfilter also instead :?: which way is better :?:

I would leave the asharp(2, 4) on the top as it is, because it applies for every frame regardless of activity, and is much better that unfilter.

-kwag

Ok, I see you already updated the Latest Script. I'm still in testing phase here, so I'll report back later! :wink:

Could anyone post a sample made with the new script? Curious :) !

audioslave go back to the Experimental Revise Prediction tread Kwag posted one there. :wink:

Better yet here http://www.kvcd.net/k19-fast-predict...-cq-63.209.mpg

Thank you, ovg64!

comparing the new script with unfilter and the old with asharp
after load in vdubmod,i see that the new loose the "dimensional"
sharpness(blurry?)... loose details,seems (is) worse.

:arrow: please,someone load the scripts in vdub
and save one bitmap for each script in the same scene to compare.

anyone see the same?

ps:
i see "marching ants"(i call it before erroneous as"staircaise")
using 480x480,but this nothing have to asharpen or unfilter,
i see it days ago and think that is the "resize" cause it.

another problem is that many people got high CQ with
544 or more resolution and i with "only" 480x480 don't get more
than CQ56 again using ToK0052 with all advices.
then something is wrong .

A stupid question. Is there any difference between the following lines:

and
???

The first line is from kwag's "Next changes to optimal script" post and the second is from the latest "Optimal script". I'm no programmer, but does the "\" symbol just tell the function to continue on the next line?

The "\" breaks long lines. In simple words, it's a "continue on next line" escape character :)

-kwag

Thank you for that info, kwag! I was just worried I had to restart my encode :) !

Jorel:

You should still use aSharp as a sharpening filter, and just use Unfitler for softening again. This actually shouldn't decrease the sharpness of the picture... :roll:

jell my friend,
did you compare the same source with pictures or samples after encode?
i did and see big differences,believe me :!:
you don't see nothing different :?:
:?

ps:
jell,i'm only trying to got the best and i test everything all day long here!
my post is nothing more to help everybody or to got help for myself.
:wink:

Jorel if you getting CQ under 60% with 480x480 than there is defenetly something wrong there, lets take a look at your script.

ok Jell,thanks

this is the script and i got the seamless Cq with asharp or unfilter:

MaxTreshold=1.50
nf=0 # Current frame.

Mpeg2Source("D:\La Luna\Segment2\DVD2AVI_PROJECT_FILE.d2v")

FieldDeinterlace()
undot()
Limiter()
asharp(2,4)
GripCrop(480,480,overscan=1,source_anamorphic=fals e)
GripSize(resizer="BicubicResize")
STMedianFilter(8,32,0,0)
MergeChroma(blur(MaxTreshold))
MergeLuma(blur(0.2))
ScriptClip("nf=YDifferenceToNext()"+chr(13)+"nf>2. 5?\
unfilter(-(fmin(round((nf/0.5)),100)),-(fmin(round((nf/0.5)),100))):\
TemporalSoften(2,7,7,3,2)")

GripBorders()
Limiter()

function fmin(float f1,float f2){return(f1<f2)?f1:f2}


anything wrong?
:?
i post some in the prediction thread too cos my CQ is too low!

thanks again Jell!
:wink:

@jorel:

I don't see any errors in that script you posted... How long is the movie?
Full Screen?
How come you don't use any overscanning on the sides at all?
For a resolution of 480x480 I can use at least Letterbox(0,0,20,20)!
That saves a lot of bitrate right there! :roll:
Another thing is that FieldDeinterlace() will give you a higher filesize too, but there's nothing you can do about that!
I just saw that your source is 29.97 fps. That makes another huge difference in Filesize.

Some movies are just not compressible at all, and you cannot really figure out why. I'd recommend you try that script on 3 more movies, and If your results are still bad, then :oops: :?: :?: :!:
We'll see...

BTW: I compared the script with aSharp softening and Unfilter softening frame by frame in VDub, and the only thing I saw was that flashes on scene changes are gone using Unfilter as a softener. Otherwise the scripts looked almost identical. I did notice that on high action scenes Unfilter blurs a lot more than aSharp. Still pictures looked identical in my eyes though. :roll:

thanks jell!
:wink:

you wrote:
Otherwise the scripts looked almost identical.
Still pictures looked identical in my eyes though.

not in mine pictures Jell,
can i send 2 pictures by mail :?:
(one with unfilter new script,another with asharp old script)
they are really differents!

and:
How long is the movie?
is Total Time : 01:28:31 fullscreen

and without deinterlace change just a little!

please,pm your mail!
:)

Yep, I agree 100% :!:
Still and low moving scenes look very sharp.
With the new values set for unfilter, we should get an average of ~(-60,-60) on heavy action scenes, and that really smoothes the picture. So we get a higher CQ values with the new revised script :)

-kwag

Hi jorel,

Can you try the script on a movie that is 23.976fps :idea: :?:

-kwag

no Kwag,not try yet but i will do it.

can i send to you the pictures from the scripts by mail?
i insist, they have BIG differences :!: :!: :!:
my work was "image quality" for 30 years,
i see big details between the scripts...
i have 5 televisions, tv out on pc (ati) and a new monitor(3 months).
:roll:

@jorel:

Hey nobody here things you cannot judge picture quality, please don't get offended my friend. :)

eMail me the pictures: catch22tx@gmx.net

I'll take a look at it too. It really might be the framerate, or the fact that the movie is fullscreen. I suppose your source is a DVD? I'm talking about the source of the screenshots you're going to send.
Regarding your other problem, your CQ seems to be right for a 90 min. FULL SCREEN movie @29.97 fps wich has to be de-interlaced.
Don't worry, I'm sure your system is ok!

of course Jell :D
i don't feel offended ...
friends have liberty to talk clear to another friend and
i don't see offence in any moment.
:wink:
for this reason i post my doubts here in forum,
not only for the results but for the great friendship.
this is the best here....you all :!:

and believe me,my friends from work call me when they have problems
with quality and when are changing the systems of tvs(ntsc to pal,to pal-n,etc).i am really "bored" with color and image quality.
the idea is get the best quality!
this is the best place to show all doubts!

and yes,my source is dvd,full screen, 29.97,ntsc,using fielddenterlace!

on "moving pictures"(fast) i see more differences.
when don't have moviments,seems the same.

sending the pictures...
thanks !
:wink: