Color banding and noise removal

[TheProfosist]

Originally Posted by DJATOM

No, I meant XamD source. It's upscaled from 540p and then OP/ED texts was applied onto it. So descaling will introduce artifacts on those captions.

Ah well I can worry about that when I get the rest of the stuff working. Right now I dont have a specific source ill be using it on. Also back then it wasnt uncommon for me to use ordered chapters (i think thats the correct term) and clean op/ed if the anime allowed for it.

Right now im mostly interested in the flow of using dither. What you would use where are why. In that previous post I gave some examples. No clue if any of that is relevant anymore either since its been so long.

[real.finder]

in dither.avsi Dither_resize16nr Function I think there are bug in this line

	sr_v  = Float (height) / Float (src.height ())

the "Float (src.height ())" should be "Float (src.height ()/2)" since it's lsb stacked clip

find it when I was porting/mixing things to support HBD

here what I have for now, will add it later to ResizeX, any suggestions?

# base on Dither_resize16nr
Function sResizeXnr (clip src, int width, int height,
\	float  "src_left",
\	float  "src_top",
\	float  "src_width",
\	float  "src_height",
\	string "kernel",
\	int    "taps",
\	float  "a1",
\	float  "a2",
\	string "cplace",
\	bool   "luma",
\	bool   "chroma",
\	bool   "mt",
\	string "mt_params",
\	string "dither_params",
\	bool   "desampling",
\	int    "dither_mode",
\	bool   "noring"
\	)
{
	noring = Default (noring, true)

	Assert (width > 0 && height > 0, "sResizeXnr: width and height must be > 0.")

	sr_h  = Float (width ) / Float (src.width () )
	sr_v  = Float (height) / Float (src.height ())
	sr_up =       max (sr_h, sr_v)
	sr_dw = 1.0 / min (sr_h, sr_v)
	sr    = max (sr_up, sr_dw)
	Assert (sr >= 1.0)

	# Depending on the scale ratio, we may blend or totally disable
	# the ringing cancellation
	thr = 2.5
	nrb = (sr > thr)
	nrf = (sr < thr + 1.0 && noring)
	nrr = (nrb) ? min (sr - thr, 1.0) : 1.0
	nrv = (nrb) ? Round ((1.0 - nrr) * 255) * $010101 : 0

	main = src.ResizeX (width, height,
\		src_left  =src_left,
\		src_top   =src_top,
\		src_width =src_width,
\		src_height=src_height,
\		kernel    =kernel,
\		taps      =taps,
\		a1        =a1,
\		a2        =a2,
\		cplace    =cplace,
\		luma      =luma,
\		chroma    =chroma,
\		mt        =mt,
\		mt_params =mt_params,
\		dither_params=dither_params,
\		desampling=desampling,
\		dither_mode=dither_mode
\	)

	nrng = (nrf) ? src.ResizeX (width, height,
\		src_left  =src_left,
\		src_top   =src_top,
\		src_width =src_width,
\		src_height=src_height,
\		kernel    ="gauss",
\		a1        =100,
\		cplace    =cplace,
\		luma      =luma,
\		chroma    =chroma
\	) : main

	nrm = (nrb && nrf) ? main.BlankClip (color_yuv=nrv, height=main.Height()) : main

	rgm  = 1
	rgc  = (nrb) ? -1 : 0
	ild  = Defined(luma)
	icd  = Defined(chroma)
	rgl  = ild ? (luma   ? rgm : rgc) : rgm
	rgch = icd ? (chroma ? rgm : rgc) : rgm
	y    = ild ? luma ? 3 : 2 : 3
	u    = icd ? chroma ? 3 : 2 : 3
	v    = u
	(nrf       ) ? main.repair (nrng, rgl, rgch)                          : main
	(nrf && nrb) ? mt_merge (main, last, nrm, y=y, u=u, v=v, luma=chroma) : last
}

[real.finder]

can we have port of vs fmtconv to avs/avs+ with HBD? I try to update/port/rewrite crt_display to avs+ HBD then I see it use fh, fv in dither_resize16

and its not only for crt_display and "fh, fv", dither_resize16/fmtconv has many other things like kernel inversion mode and additional Kernels

[tormento]

Originally Posted by real.finder

can we have port of vs fmtconv to avs/avs+ with HBD?

AVSResize wouldn't fit?

[real.finder]

Originally Posted by tormento

AVSResize wouldn't fit?

it's not https://forum.doom9.org/showthread.p...54#post1906754

[cretindesalpes]

Dither 1.28.0:

• Added Dither_lutspa8 and Dither_lutspa16.
• Fixed height issue in Dither_resize16nr. Thanks to real.finder for the report.
• DLL recompiled with Visual C++ 2019.

[kedautinh12]

Originally Posted by cretindesalpes

Dither 1.28.0:

• Added Dither_lutspa8 and Dither_lutspa16.
• Fixed height issue in Dither_resize16nr. Thanks to real.finder for the report.
• DLL recompiled with Visual C++ 2019.

Thanks

[real.finder]

Originally Posted by cretindesalpes

Dither 1.28.0:

• Added Dither_lutspa8 and Dither_lutspa16.
• Fixed height issue in Dither_resize16nr. Thanks to real.finder for the report.
• DLL recompiled with Visual C++ 2019.

thanks, but is there any plan to support HBD in avs+ or port fmtconv to avs? since the HBD filters nowadays are way more than lsb ones and also there are float point in HBD of avs+

[cretindesalpes]

real.finder: sorry, there is no plan at the moment. I don’t do much video processing these days and have other projects with higher priority.

EDIT—another update:

MVTools 2.6.1.0

• MSuper, MDegrain1/2/3/N: YV16 and YV24 colorspaces are supported for filtering. Analysis still requires YV12 or YUY2 clips.
• FlowFPS, FlowInter and MMask: fixed a memory intialization bug that could produce slightly different results each run. Thanks to dubhater for the fix.
• MAnalyse, MRecaculate: fixed block size parameter check in divide mode. Thanks again to dubhater.
• MDepan: parameter zerow is now actually taken into account.
• DLL recompiled with VC++ 2019, source files and projects modified accordingly.

EDIT—second update:

Dither was updated to v1.28.1. Fixed a regression in mode 20 of Dither_removegrain16.

[tormento]

Sorry to resurrect some old stuff but...

Is there any newer version of Dither?

I have checked 1.28.1 version and it's still a 2.5 plugin.

If positive, perhaps I have lost the release. Please address me

[real.finder]

Regarding Dither modernization for HBD, I think it's impossible to do it as it now, After thinking and talking with people that do encode from long times, it's better to port dither_bilateral16 (it's seems work differently from the rest and as some friend said "it strikes a nice balance between strength and detail loss"), Dither_box_filter16 (same as dither_bilateral16 case), and SmoothGrad to independent plugins and backport fmtconv (ported done https://forum.doom9.org/showthread.php?t=183139 thanks to cretindesalpes), if these done then we will not need dithertools anymore in avs+

[kedautinh12]

True, but no one backport it now

[kedautinh12]

I seen 299792458m update dfttest 1.9.6 of pinterf with new commits, can anyone compile it?? thanks
https://github.com/299792458m/dfttest_mod/commits/196

[kedautinh12]

New fork of fmtconv
https://github.com/AkarinVS/fmtconv

[cretindesalpes]

Originally Posted by real.finder

backport fmtconv [...] in avs+

I’m currently working to make fmtconv easier to port to other standards. The bitdepth part requires the most work here. However I’m not sure if I’ll do the final port yet.

[StainlessS]

Originally Posted by cretindesalpes

I’m currently working to make fmtconv easier to port to other standards. The bitdepth part requires the most work here. However I’m not sure if I’ll do the final port yet.

The pussycat is still alive, we are all happy to hear that,
live long and prosper CD. [OK you made some recent posts too, I did not miss all of them].

[GMJCZP]

Originally Posted by StainlessS

The pussycat is still alive, we are all happy to hear that,
live long and prosper CD. [OK you made some recent posts too, I did not miss all of them].

+1, Thanks cretindesalpes for your contributions.

[real.finder]

Originally Posted by cretindesalpes

I’m currently working to make fmtconv easier to port to other standards. The bitdepth part requires the most work here. However I’m not sure if I’ll do the final port yet.

good news, thanks! if you have time can you port dither_bilateral16, Dither_box_filter16 and SmoothGrad to independent plugins with new appropriate names and HBD?

[kedautinh12]

DFTTest latest ver
https://github.com/pinterf/dfttest/releases

[kedautinh12]

Updated dither - 1.28.1.1 x64 from Asd-g: updated to 2.6 plugin, added support for passthrough frame properties
https://gitlab.com/uvz/AviSynthPlus-...%2B/dither.dll