New optimal script (V4) is there
 

Hi all,

After many tests and some good feedbck about it, this is the new optimal script for making avi->kvcd encoding :
The second Blockbuster line should be used in rare occasions when there is too much DCT blocks done by the MPEG encoder. I used it only twice.

So the only difference is that ATC is subtituted with Convolution3d, right?

Right.

Why you choose 2 times Blockbuster and not deen? With deen I have very good results at a very low bitrate and old video better tha blockbuster.

What's your point Encoder Master?
You are comparing a Denoiser with a filter to stray noise in low detail areas... How can you exchange one for the other?

Oh sorry I had a wrong knowledge about Blockbuster and I 've thought It's like deen. :oops: :oops: :oops:

And you had bad knowledge also of the script history : V1 used Deen !

We dropped it because it does not do a good job at all.

Was your message just for giving us a chance to see again your boring signature ? Wa are not there to make silly adds for websites ! :-(

Thank you very much for the update Dialhot ive been using the newer V2 script and had some excellent results look forward to seeing the improvements with this one. :)

Encoder Master ...
My recommendation:

1. Read the threads!... especially in here, the Avisynth- and Optimal Script section! Read the threads of the released optimal scripts 1,2,3 and you will see why there is something used in a line of a script.

2. Read the readme's of each filter you downloaded! As I recommended it to you already.

3. Get rid of your SIG. It has nothing to do with mpeg business or with contents about yourself

First attempt with the new script and this is the error i get

First time i have ever got this message

If all i have to add is a convert toYUY2 line where do ihave to put it

thanx

you don't have to use ConverttoYUY2. There is a convolution3d plugin made for YV12 colorspace, so use that instead. :)

THnax muhali3,where do i get it ,i just downloaded the convolution 3d today from warpenterprises.did i get the wrong one

Maybe you already know this but the temporal influence is disabled in convolution3dyv12.

It should be a funny game. :cry:
And sorry with the critik wasn't correct. :wink:

Yes, but we need the temporal influence there.
Try without it to see the difference.

Which means that you should use the YUY2 version in your script. The YV12 version doesn't do any temporal filtering as Dano stated.

sorry guys, my mistake. :roll: :lol:

I didn't notice that. So I change my answer : I din't know what they called "temporal influence" in the doc, but it seems that the temporal _filtering_ is on.

Influence is normally the forth paremeter (set to 2.8 in the script). I guess is something used for balancing temporal filtering according to some paramaters, but really don't see anything else.

Now, if I'm wrong, then that means that the script do not need a spatio-temporal filter there, but I'm pretty sure it does (I tried other filters and spatial only ones gave poor results).

Ok guys,is the solution to my problem in those answers,im having trouble figuring it out

Thanx again

Dialhot please don't yell at me if it is :D because :imstupid:

well bigggt, im using the YV12 version and i'm still getting good results. I am encoding the matrix (1st one) and the quality is very good at 352x480, the only problem is the mosquito noise. It's all over the place :x

could someone suggest what i should do to reduce mosquito noise at low resolutions (<352x480)?

Thanks :)

Thanx muhali3,just got this version and it seems to be working

Convolution3D-YV12 released on 03-29-2003

Is this right

On a side note when you guys talk about noise,mosquito noise is there somewhere that has picture of all these things so i can try and learn whatyou guys are talking about.

yes that is the right convolution 3d plugin.

mosquito noise is noise that appears in areas with high contrast. Such as white lettering on a black background, like in the end credits. In the credits, if there was mosquito noise, it would be surrounding the outside of the letters and other places with high contrast. It is very noticeable at low resolutions.

I can't find a thread with a picture. I have just learned it from reading the threads.

Hope i answered your question. :o

Did you check the AVI? Is there already mosquito noise? Is it already very sharp? Did you apply another SharpenFilter? Because sharpen edges are most responsable for mosquito noise. So do a test and change the word "Lanczos" into "Bicubic" in the GripfitResizer. And maybe the mosquito noise is gone but your movie will come out sharp enough.

Encoding Matrix no matter if part 1 or 2 will give you a real high CQ! and therefore almost no mosquito noise, cause the movie is 50% dark and therefore it loves to be compressed.

The best you can do is posting a pic of before and after in here so we can see what would be the best. ;-)

no there is not mosquito noise. its not already sharp. and i didn't put any sharpen filters. what i see in the source is dct-blocks and rainbows (real bad divx).

i'll try to get some pics
[/code]

Sorry my friend but it seems you go "off topic" by exactly using this script ;-)
Convo3d set to "presets" and TempCleaner do not belong to the script to which this thread is refering to.
So how about using the script as mentioned in the first post of this thread?
Give it a try ... and then we can go further with optimizations refering to your special source condition. ;-)
BTW: Always put the BlindPP(..) line next to the Avisource(...) line cause that's what it does .. preprocessing. And my recommendation is not to use the "movieLQ" preset in Convo3d cause it will seem like if you pour concrete on still existing details 8O

k, i just tried to change a little things in the script. thanks for the advice. i'll change it back. and see the results. :o

Good advice from Incredible but just one thing : source_ana=true AND dest_ana=true leads to do.. nothing !

In fact as you are using avi and ALL divx are non anamorphic, that is source_ana=false that you have to use and add dest_ana=true only if your TV set is widescreen.

remember 16:9 and anamorphic ARE NOT THE SAME THING !

Still eagles eyes that late at night! 8O :)

yeah, i found that out after my test encode,

@incredible: you were very right. when i encoded the matrix at 352x480, CQ 76.43, the output file was only 610MB.

:) The same will result when encoding "Bladerunner", "Matrix Reloaded!!", "Seven" .... and much more .... they all contain more than 50% Dark within the movie = fantastic CQ values!

BTW: When getting more CQ than 75 after prediction you also can add a second language Audio track .. as I did when encoding Matrix Reloaded @ mpeg1/480x576/300min/2000max the CQ was still 62 incl. 2 Audio Tracks each @ 112kbit and the picture was very good on TV. 8O. The same was possible with "Seven".

postprocessing, not preprocessing.
But you have to put if before the resizing because its purpose is to break the border of DCT blocks. These borders are very distorted by the resizing then blindPP can't process on them.

I always confound that, although I learned a little of latin in school :lol:
Preprocessing would be illogical.

f

i fixed my mosquito noise and rainbow problem. i just used GuavaComb and it cleaned it up pretty well, and thanks to incredible, blindpp next to avisource cleaned up my dct-blocks.

v4 looks great. thanks.

Maybe I've missed something but what exactly is better with this new version, comparing v3 and v4 :oops:

//Wolfi

i guess it's more compression, while keeping a good amount of details.

You guess right. But the script is still evoluting.

http://www.kvcd.net/forum/viewtopic.php?p=51400#51400