real.finder's Avisynth Stuff

[kuchikirukia]

Unfiltered: 30MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true) = 21MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true, plane=0) = 28MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true, plane=3) = 81MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true, plane=1) = 83MB

Umm....

eh, 3.1.2d does it too.

[real.finder]

Originally Posted by kuchikirukia

Unfiltered: 30MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true) = 21MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true, plane=0) = 28MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true, plane=3) = 81MB
smdegrain(thsad=200,lsb=true,lsb_in=false,lsb_out=true, plane=1) = 83MB

Umm....

eh, 3.1.2d does it too.

it's something in mvtools or the way of how the encoder work

[real.finder]

days ago I did update stab3, now it's should be more stable with mirror=15

[tormento]

I want to ask you a curiosity: is SMDegrain written to work with the 16 bit depth capabilities of MaskTool, MVTool and KNLMeansCL or does it use some kind of workaround?

AFAIK they are the only filters I use with the following:

PHP Code:

SMDegrain (tr=4,PreFilter=4,thSAD=600,contrasharp=false,refinemotion=false,truemotion=true,plane=4,chroma=true,lsb=true,mode=0) 


If not, would be a "pure" 16 bit version faster and/or more precise than the workaround method, if any?

[real.finder]

Originally Posted by tormento

I want to ask you a curiosity: is SMDegrain written to work with the 16 bit depth capabilities of MaskTool, MVTool and KNLMeansCL or does it use some kind of workaround?

AFAIK they are the only filters I use with the following:

PHP Code:

SMDegrain (tr=4,PreFilter=4,thSAD=600,contrasharp=false,refinemotion=false,truemotion=true,plane=4,chroma=true,lsb=true,mode=0) 


If not, would be a "pure" 16 bit version faster and/or more precise than the workaround method, if any?

in avs+ native 16 bit? it not use workaround, but the lsb in your code is not, it should show you some error message that lsb it not work with native HBD

edit: it will not be faster in native 16 unless you did this workaround

#8bit source
c=last

SMDegrain(globals=3,tr=4,PreFilter=4)
global fv1 = fv1.MScalevect(bits=16)
global fv2 = fv2.MScalevect(bits=16)
global fv3 = fv3.MScalevect(bits=16)
global fv4 = fv4.MScalevect(bits=16)
global bv1 = bv1.MScalevect(bits=16)
global bv2 = bv2.MScalevect(bits=16)
global bv3 = bv3.MScalevect(bits=16)
global bv4 = bv4.MScalevect(bits=16)

c
ConvertBits(16)
SMDegrain(globals=1,tr=4,thSAD=600,contrasharp=false,refinemotion=false,truemotion=true,plane=4)
ConvertBits(8,dither=1)

[tormento]

Originally Posted by real.finder

in avs+ native 16 bit?

Thanks for your clear reply.

To clarify my question: will be possible one day to have a SMDegrain (bitdepthworkingspace=8/10/16,outputbitdepth=8/10/16) without the need of LSB? Would like also to try if a pure 16 bit processing would be more precise than a "fake" one.

[real.finder]

Originally Posted by tormento

Thanks for your clear reply.

To clarify my question: will be possible one day to have a SMDegrain (bitdepthworkingspace=8/10/16,outputbitdepth=8/10/16) without the need of LSB? Would like also to try if a pure 16 bit processing would be more precise than a "fake" one.

it's already do that but there are some filters that not work with native HBD yet (in your case it support native 10-16 bit now but you should remove the "lsb=true" or make it false), but that is way slower than the lsb method, even with the code I post above, see here https://forum.doom9.org/showthread.p...87#post1833887

[tormento]

@real.finder

Could you please tell me which is the difference between

PHP Code:

pre=KNLMeansCL (d=0, a=3, h=7.0)
SMDegrain (tr=6, thSAD=400, refinemotion=false, lsb=true, mode=0, contrasharp=false, PreFilter=pre, truemotion=false, plane=4, chroma=true) 


and

PHP Code:

pre=SMDegrain_KNLMeansCL (d=0, a=3, h=7.0)
SMDegrain (tr=6, thSAD=400, refinemotion=false, lsb=true, mode=0, contrasharp=false, PreFilter=pre, truemotion=false, plane=4, chroma=true) 


They are giving me different results.

[real.finder]

Originally Posted by tormento

@real.finder

Could you please tell me which is the difference between

PHP Code:

pre=KNLMeansCL (d=0, a=3, h=7.0)
SMDegrain (tr=6, thSAD=400, refinemotion=false, lsb=true, mode=0, contrasharp=false, PreFilter=pre, truemotion=false, plane=4, chroma=true) 


and

PHP Code:

pre=SMDegrain_KNLMeansCL (d=0, a=3, h=7.0)
SMDegrain (tr=6, thSAD=400, refinemotion=false, lsb=true, mode=0, contrasharp=false, PreFilter=pre, truemotion=false, plane=4, chroma=true) 


They are giving me different results.

of course they are different, since SMDegrain_KNLMeansCL process chroma with luma by special way

[tormento]

Originally Posted by real.finder

of course they are different, since SMDegrain_KNLMeansCL process chroma with luma by special way

Please define "special".

[real.finder]

Originally Posted by tormento

Please define "special".

back then and still KNLMeansCL can't process chroma with luma in subsampling Chroma (new KNLMeansCL can do process chroma alone) so dogway use stacking method, you can see the code yourself, or back to dogway posts in videohelp forum

[magiblot]

Hi real.finder,

Is there any way to run srestore in AVS+ x64? At first sight, it won't run because there doesn't seem to be a x64 version of GScript. Since I have read AVS+ supports GScript's functionality natively, I tried to replace the call to GScriptClip in srestore.avsi with ScriptClip, but a "I don't know what 'dm' means" error is the farthest I got.

Thanks.

[StainlessS]

GScriptClip is from Grunt plugin, not GScript.

[magiblot]

Wow, what a miss.
Thanks, Stainless .

[real.finder]

Originally Posted by magiblot

Hi real.finder,

Is there any way to run srestore in AVS+ x64? At first sight, it won't run because there doesn't seem to be a x64 version of GScript. Since I have read AVS+ supports GScript's functionality natively, I tried to replace the call to GScriptClip in srestore.avsi with ScriptClip, but a "I don't know what 'dm' means" error is the farthest I got.

Thanks.

as StainlessS said, it's Grunt

and I think the error related to something else (not Grunt)

you use mt "Prefetch()"? cuz it will not work, runtime things (srestore use it) not compatible with mt yet

[tormento]

Originally Posted by real.finder

back then and still KNLMeansCL can't process chroma with luma in subsampling Chroma (new KNLMeansCL can do process chroma alone) so dogway use stacking method, you can see the code yourself, or back to dogway posts in videohelp forum

Looked.

What does this line

!lsb && lsb_in ? Ditherpost(mode=6,slice=false) : last

means?

As I am using SMDegrain_KNLMeansCL as a prefilter (just to have more control over prefilter parameters), is it useful to dither when SMDegrain already dither again at some point?

Wanting to use KNLMeansCL as a plain prefilter, would it be sufficient to use:

pre =KNLMeansCL (d=0, a=3, h=7.0, channels="Y")
pre2=KNLMeansCL (d=0, a=3, h=7.0, channels="UV")
SMDegrain (tr=6, thSAD=400, refinemotion=false, lsb=true, mode=0, contrasharp=false, PreFilter=pre2, truemotion=false, plane=4, chroma=true)

or the very complex prefilter script is necessary?

[real.finder]

remember that I told you that mvtools don't has lsb_in anywhere, it's just has lsb parameter in mdegrainX, so you must dither any stacked input clip

yes, you can use

pre =KNLMeansCL (d=0, a=3, h=7.0, channels="Y")
pre2=KNLMeansCL (d=0, a=3, h=7.0, channels="UV")
SMDegrain (tr=6, thSAD=400, refinemotion=false, lsb=true, mode=0, contrasharp=false, PreFilter=pre2, truemotion=false, plane=4, chroma=true)

or anything give you what you want

but I think you need to added "pre." before KNLMeansCL in pre2, so it should be

pre2=pre.KNLMeansCL (d=0, a=3, h=7.0, channels="UV")

[tormento]

Originally Posted by real.finder

remember that I told you that mvtools don't has lsb_in anywhere, it's just has lsb parameter in mdegrainX, so you must dither any stacked input clip

We need some Pinterf modding here

Originally Posted by real.finder

or anything give you what you want

[real.finder]

Originally Posted by tormento

We need some Pinterf modding here

why? lsb is hack, he already did support native avs+ HBD, and he will add out16 as replacement for lsb method, see last posts in his mvtools Thread

[tormento]

Originally Posted by real.finder

why? lsb is hack, he already did support native avs+ HBD, and he will add out16 as replacement for lsb method, see last posts in his mvtools Thread

I saw and I know. Look at smiley.