Dehalo_Beta_Package: a fork of Dehalo_alpha and FineDehalo

[DonCanjas]

https://github.com/DonCanjas/Avisynt...o_Beta_Package

Dehalo_Beta_Package consists of Dehalo_Beta, a fork of Dehalo_alpha; and FineBeta, a fork of FineDehalo.

It currenlty supports YUV[A] & Y 8-32 bit.



Dehalo_Beta

Instead of built-in scalers, Dehalo_Beta uses avsresize and has the option to use libplacebo_resample with useplacebo=True.

DGSource("haloed_clip.dgi")
ConvertBits(16)

src = last

DA = src.DeHalo_Alpha(darkstr=0)

DB = src.DeHalo_Beta()

FineBeta

FineBeta is basically the same as FineDehalo, but uses Dehalo_Beta and by default it uses Kirsch to create an edgemask.
On top of that, you can choose between 3 different masktypes: kirsch, old, TEdge. Old uses FineDehalo's default mask, TEdge will use TEdgemask and kirsch uses ex_edge's kirsch.

You can also parse your own edgemask to FineBeta and, as in FineDehalo, input your own halo mask.

DGSource("haloed_clip.dgi")
ConvertBits(16)

src = last

FDH = src.FineDehalo(darkstr=0)

FB = src.FineBeta()

I'm planning on adding chroma dehaloing support for FineBeta. I'm thinking on using SSIM_downsample() to downscale luma's mask to chroma size and process every plane individually.

[real.finder]

so DeHalo_Beta should give same output as DeHalo_alpha_2BD?

[DonCanjas]

Originally Posted by real.finder

so DeHalo_Beta should give same output as DeHalo_alpha_2BD?

I forked it from Dehalo_alpha_mt2. It should have a similar output, tho you can see that it's not 1:1 when working with 16 bit video

[coolgit]

Here lies a problem.

Every time someone comes up with a plugin/script/whatever and post images showing suppose differences, I can't see where the differences are.

Any chance you could insert and arrow or something onto the images to show what the differences are?

[StainlessS]

CoolGit,

Maybe something like this [flip from even to odd frames]

DGSource("haloed_clip.dgi")
ConvertBits(16)

src = last

DA = src.DeHalo_Alpha(darkstr=0)

DB = src.DeHalo_Beta()

Interleave(DA,DB)  # Even DA, Odd DB

Or similar for any other test script

EDIT: OR eg,

DGSource("haloed_clip.dgi")
ConvertBits(16)

src = last

DA = src.DeHalo_Alpha(darkstr=0)

DB = src.DeHalo_Beta()

Interleave(src.Subtitle("Src"), DA.Subtitle("DA"), DB.Subtitle("DB"))

Untested.
EDIT: or with Src adjacent to both DA and DB [group of 4]

Interleave(src.Subtitle("Src1"), DA.Subtitle("DA"), DB.Subtitle("DB"), src.Subtitle("Src2"))

Or,

StackVertical(StackHorizontal(src.Subtitle("Src"),DA.Subtitle("DA")),StackHorizontal(src.Subtitle("Src"),DB.Subtitle("DB")))

Or,

StackVertical(StackHorizontal(src.Subtitle("Src"),DA.Subtitle("DA")),StackHorizontal(Subtract(DA,DB).Subtitle("DA-DB"),DB.Subtitle("DB")))

Again Untested.

[coolgit]

How do I do that? I don't see haloed_clip.dgi uploaded anywhere.

[StainlessS]

I was just using example script in opening post,

DA, and DB, could be any 2 alternative results of any script.

[takla]

Both give me this error:
There is no function named "m4_sh". Line 31 & 33

[DonCanjas]

Originally Posted by takla

Both give me this error:
There is no function named "m4_sh". Line 31 & 33

I forgot to list Zs_RF_Shared.avsi as required. Should work if you download it:
https://github.com/realfinder/AVS-St...RF_Shared.avsi

[tormento]

Originally Posted by DonCanjas

Dehalo_Beta_Package consists of Dehalo_Beta, a fork of Dehalo_alpha; and FineBeta, a fork of FineDehalo.

Would be possible to have a updated version to work with AVS+ without the need of stacking?

Plus I can't find Retinex_MSRCP anywhere.

Thank you!

[poisondeathray]

Originally Posted by tormento

Would be possible to have a updated version to work with AVS+ without the need of stacking?

Did you try this version ?
https://github.com/realfinder/AVS-St...alpha_mt2.avsi

# DeHalo_alpha with high bit support, need masktools 2.2.20 or up in avs26 and last avs+
# 2022.08.31

Plus I can't find Retinex_MSRCP anywhere.

https://github.com/Asd-g/AviSynth-Retinex

MSRCP (clip clip, float[] "sigma", float "lower_thr", float "upper_thr", bool "fulls", bool "fulld", float "chroma_protect")

[tormento]

Originally Posted by poisondeathray

Did you try this version?

Actually, it's one of the various dehalo that I tried.

FineBeta is way more effective and less prone to destroy details.

Originally Posted by poisondeathray

Retinex

I had to rename RETINEX_MSRCP to MSRCP in the script to make it work.

[poisondeathray]

Originally Posted by tormento

I had to rename RETINEX_MSRCP to MSRCP in the script to make it work.

You should let the author know so he can update on github

[tormento]

Originally Posted by poisondeathray

You should let the author know so he can update on github

Did it.

[tormento]

Whenever I flag useplacebo=true, I get:

libplacebo_Resample: [error] vk->CreateDevice(vk->physd, &dinfo, PL_VK_ALLOC, &vk->dev): VK_ERROR_INITIALIZATION_FAILED (../src/vulkan/context.c:1299)
[fatal] Failed creating logical device!
[fatal] Failed initializing vulkan device

Avslibplacebo works fine as standalone filter on my computer.

Any idea?