TComb for VapourSynth

[jackoneill]

https://github.com/dubhater/vapoursynth-tcomb/releases

This is a port of TComb v2.0 beta 2 for Avisynth.

The "opt" parameter is no more. The other parameters should all behave the same.

If someone really needs 16 bit support, let me know and I may look into that.

[mastrboy]

Thanks.

For 16bit, It's always useful to have so you don't have to go back and forth between 8/10/16 bit, I always prefer to have my entire filter chain at the highest bit level supported.

[jackoneill]

Thanks.

For 16bit, It's always useful to have so you don't have to go back and forth between 8/10/16 bit, I always prefer to have my entire filter chain at the highest bit level supported.

This is supposed to go right after the source filter, so unless you have a high bit depth source, I won't bother.

[Reel.Deel]

A true high bit depth source with dot crawl/rainbows is probably as rare as they come. The only dot crawl and rainbows I've ever encountered are on DVDs. On another note, the documentation says cross-luminance (rainbowing) and cross-chrominance (dot crawl) but shouldn't it be the other way around?

[feisty2]

there're some cheesy mastertapes with this crap also.
and 16bpc could be useful,
sometimes a single method doesn't work so well on composite videos
there're some other methods like "MDegrain" approach and "Frfun7" approach work pretty well as well
if you want to merge the results of these methods, 16bpc is useful, since MDegrain is 16bpc

[jackoneill]

On another note, the documentation says cross-luminance (rainbowing) and cross-chrominance (dot crawl) but shouldn't it be the other way around?

Uh, no idea.

[zerowalker]

indeed it should. Luminance brings dot-crawl while Chroma brings rainbowing (quite hard for Luma (Black/White) to make a rainbow;P).

[jackoneill]

indeed it should. Luminance brings dot-crawl while Chroma brings rainbowing (quite hard for Luma (Black/White) to make a rainbow;P).

Actually no, luminance does not bring dot crawl. Dot crawl is found in the luma, but it's caused by a bit of the chroma leaking into the luma. (Rainbows are caused by a bit of luma leaking into the chroma.)

[feisty2]

cross luminance = crossing luminance = high frequent luma information is interfering chroma channels = rainbow
same goes for dot crawl, that's my guess

[Reel.Deel]

Did a little bit of research. From the Doom9 guides (12.3 Removing rainbow effects and dot crawls (www.doom9.org/index.html?/capture/chroma_artefacts.html)): "The technical terms are as follows: rainbow effects is cross color (hi-frequency luma data upsets the chroma demodulator) and dot crawl is cross luminance (chroma leftovers in the Y signal)". Another reference is from the book CCTV: Networking and Digital Technology: "Chroma crawl. An artifact of encoded video, also known as dot crawl or cross-luminance. Occurs in the video picture around the edges of highly saturated colors as a continuous series of crawling dots and is a result of color information being confused as luminance information by the decoder circuits." I also translated CC (http://avisynth.nl/index.php/CC), DeCross (http://avisynth.nl/index.php/DeCross), and DeDot (http://avisynth.nl/index.php/DeDot) and they also use the same terminology. Really TComb is the only one that uses this terminology backwards, it made me wonder when I added it to the wiki some time ago.

[jackoneill]

Fixed.

[jackoneill]

Unexpectedly, v2 is here. (https://github.com/dubhater/vapoursynth-tcomb/releases/tag/v2)

* Fix a bug in the asm

There is a 64 bit Linux binary now.

[KingLir]

Hi all, I am unable to build on macOS (latest 10.12.3), failed on make:

Undefined symbols for architecture x86_64:
"_tcomb_horizontalBlur6_c", referenced from:
_HorizontalBlur6 in tcomb.o

And here is the entire console log:
https://mega.nz/#!Ik4VUYqT!Wg6Yc8by-Az9cDb-k_3ntw1n6DRbhSsOprBqPEs-GEA

Anyone can help/fix ? Or if you already have .dylib build for macOS - can you share it with me ?

[jackoneill]

Hi all, I am unable to build on macOS (latest 10.12.3), failed on make:

Undefined symbols for architecture x86_64:
"_tcomb_horizontalBlur6_c", referenced from:
_HorizontalBlur6 in tcomb.o


I have no idea what's wrong there.

[jackoneill]

On second thought, try this patch: https://dpaste.de/FDdc/raw

[KingLir]

On second thought, try this patch: https://dpaste.de/FDdc/raw

Sorry for the newbie question, but I download zip from GitHub "Clone or Download" green button.
Then executed the following on terminal (1.diff holds all 193 lines from your patch):
git apply 1.diff

I am getting this:
warning: src/tcomb.c has type 100755, expected 100644
error: patch failed: src/tcomb.c:76
error: src/tcomb.c: patch does not apply

It seems the end of the patch is missing (should be more than 193 lines).

[jackoneill]

193 lines is exactly right.

This should fix the error:

chmod -x src/tcomb.c


To avoid errors like that in the future download the tar.gz instead of the zip, or use git. Edit: I see the Clone or download button doesn't have a tar.gz option. In this case, you can just download the v2 release as tar.gz. It's the same code you got already.

[KingLir]

193 lines is exactly right.

This should fix the error:

chmod -x src/tcomb.c


To avoid errors like that in the future download the tar.gz instead of the zip, or use git. Edit: I see the Clone or download button doesn't have a tar.gz option. In this case, you can just download the v2 release as tar.gz. It's the same code you got already.

Tried and with both I am getting now:
error: patch failed: src/tcomb.c:76
error: src/tcomb.c: patch does not apply

[jackoneill]

Then just read the patch and apply the changes manually with your favourite text editor.

[KingLir]

Then just read the patch and apply the changes manually with your favourite text editor.

Sorry, I don't know how to do that without mistakes.

I also tried SourceTree "Apple Patch" but everything is detected as "rejected hunks".
May you just do a commit ? And if for some reason you won't be happy with it in the future - do a revert.

[Are_]

Did you try it? You only have to add the word "static" at the beginning of the lines the patch tells you.

[KingLir]

Did you try it? You only have to add the word "static" at the beginning of the lines the patch tells you.

Thanks! everything works great now.
jackoneill - you can safely do that commit :)

[jackoneill]

Okay, I pushed a commit.

[jackoneill]

v3 is here (https://github.com/dubhater/vapoursynth-tcomb/releases/tag/v3).


* Fix small rounding error in the asm (bug introduced in v1).
* 32 bit builds may be faster now, about as fast as the Avisynth plugin, maybe.


The rounding error probably didn't change the output visibly.

I converted the external asm into intrinsics, because there are other things I should be doing.

[KingLir]

v3 is here (https://github.com/dubhater/vapoursynth-tcomb/releases/tag/v3).


Thanks! Here is the macOS build - you can add it to the "releases" on GitHub.

https://mega.nz/#!1hwVmTLI!fS0nznbJX1TB7QqcdSuWh2DVuwRrano6d6hRPFvdPLs

[jackoneill]

Thanks! Here is the macOS build - you can add it to the "releases" on GitHub.

https://mega.nz/#!1hwVmTLI!fS0nznbJX1TB7QqcdSuWh2DVuwRrano6d6hRPFvdPLs

Thanks, but I'd rather not upload anything I didn't compile myself.