Uband - debanding flat areas only

[anton_foy]

Here is a little function I put together that I use all the time.

# Uband - debanding script based on neo_f3kdb to affect only "flat / non detailed" areas
# Requirements: neo_f3kdb, fft3dfilter, ex_tools, masktools2, sysinfo

function Uband(clip c, float "loC", int "grainY", int "grainC", bool "mt"){

loC = default( loC, 3.0 ) # smoothing strength of chroma in flat areas

# Neo_f3kdb parameters:
grainY = default( grainY, 23 )
grainC = default( grainC, 3 )
mt = default( mt, false )

mult= BitLShift(1,c.BitsPerComponent - 8)
nc = si_physicalcores()
O = c.bicubicresize(c.width/3,c.height/3).ex_edge(scale=3,"qprewitt").ex_expand().ex_invert().levels(0,0.1,212*mult,0,255*mult).bicubicresize(c.width(),c.height())
N = c.fft3dfilter(sigma=0,sigma2=0,sigma3=0,sigma4=loC,bt=1,plane=3,ncpu=nc).neo_f3kdb(grainY=grainY,grainC=grainC,mt=mt)
M = mt_merge(c,n,o,luma=true,Y=3,U=3,V=3)
S = c.bicubicresize(c.width/4,c.height/4).greyscale().levels(36*mult,0.16,70*mult,0,100*mult).ex_invert().bicubicresize(c.width(),c.height())
mt_merge(c,m,s,luma=true,Y=3,U=3,V=3)
return last}

Maybe useful for someone.

[guest]

Here is a little function I put together that I use all the time.

# Uband - debanding script based on neo_f3kdb to affect only "flat / non detailed" areas
# Requirements: neo_f3kdb, fft3dfilter, ex_tools, masktools2, sysinfo

function Uband(clip c, float "loC", int "grainY", int "grainC", bool "mt"){

loC = default( loC, 3.0 ) # smoothing strength of chroma in flat areas

# Neo_f3kdb parameters:
grainY = default( grainY, 23 )
grainC = default( grainC, 3 )
mt = default( mt, false )

nc = si_physicalcores()
O = c.bicubicresize(c.width/3,c.height/3).ex_edge(scale=3,"qprewitt").ex_expand().ex_invert().levels(0,0.1,212*256,0,255*256).bicubicresize(c.width(),c.height())
N = c.fft3dfilter(sigma=0,sigma2=0,sigma3=0,sigma4=loC,bt=1,plane=3,ncpu=nc).neo_f3kdb(grainY=grainY,grainC=grainC,mt=mt)
M = mt_merge(c,n,o,luma=true,Y=3,U=3,V=3)
S = c.bicubicresize(c.width/4,c.height/4).greyscale().levels(36*256,0.16,70*256,0,100*256).ex_invert().bicubicresize(c.width(),c.height())
mt_merge(c,m,s,luma=true,Y=3,U=3,V=3)
return last}

Maybe useful for someone.

Hi anton,

I got this to work for me, but it seems to be fairly particular to the type of video I want to "Uband".

There is a problem with line 11, on some & line 14 on others.

Not sure why.

[kedautinh12]

If you use ripbox which don't alarm what error exactly. Waste time for you and me to fix it

[guest]

If you use ripbox which don't alarm what error exactly. Waste time for you and me to fix it

Now come on ked, it's RipBot !!!! ;)

Well, RipBot uses AVSMeter to check scripts, so unless there is another way to use AVSMeter so it does show what error, then that's all I can do, as I don't use any other app's !!!

I occasionally use Staxrip & Handbrake.

[anton_foy]

Hi anton,

I got this to work for me, but it seems to be fairly particular to the type of video I want to "Uband".

There is a problem with line 11, on some & line 14 on others.

Not sure why.

Oh its hard for me to know which kind of error. 11 you say, that is"mt" and default is false so it is strange it shouldn't be a problem. Did you have all the required dlls?

[guest]

Oh its hard for me to know which kind of error. 11 you say, that is"mt" and default is false so it is strange it shouldn't be a problem. Did you have all the required dlls?

Hi anton,

It just weird how it works on some video's but not on others, and I don't really know why.

Is there a particular "type" (colour space, etc) that it requires ??

But if you go here (where I answered another users question), you'll see my script (the bottom one) :-

https://forum.doom9.org/showthread.php?p=1975088#post1975088

[StainlessS]

Is there a particular "type" (colour space, etc) that it requires ??

O = c.bicubicresize(c.width/3,c.height/3).ex_edge(scale=3,"qprewitt").ex_expand().ex_invert().levels(0,0.1,212*256,0,255*256).bicubicresize(c.width(),c.height())
N = c.fft3dfilter(sigma=0,sigma2=0,sigma3=0,sigma4=loC,bt=1,plane=3,ncpu=nc).neo_f3kdb(grainY=grainY,grainC=grainC,mt=mt)
M = mt_merge(c,n,o,luma=true,Y=3,U=3,V=3)
S = c.bicubicresize(c.width/4,c.height/4).greyscale().levels(36*256,0.16,70*256,0,100*256).ex_invert().bicubicresize(c.width(),c.height())

I have not really looked too hard at it, but all of the "*256" stuff makes it seem 16 bit specific to me.


8 bit *1, # * [ 2 ^ (8-8) ] ::: * BitLShift( 1 , 8-8)
10 bit *4, # * [ 2 ^ (10-8) ] ::: * BitLShift( 1 , 10-8)
12 bit *16, # * [ 2 ^ (12-8) ] ::: * BitLShift( 1 , 12-8)
14 bit *64, # * [ 2 ^ (14-8) ] ::: * BitLShift( 1 , 14-8)
16 bit *256. # * [ 2 ^ (16-8) ] ::: * BitLShift( 1 , 16-8)

BPC = BitsPerComponent(Src)
MULT = BitLShift( 1 , BPC - 8 ) # Multiplier

[anton_foy]

O = c.bicubicresize(c.width/3,c.height/3).ex_edge(scale=3,"qprewitt").ex_expand().ex_invert().levels(0,0.1,212*256,0,255*256).bicubicresize(c.width(),c.height())
N = c.fft3dfilter(sigma=0,sigma2=0,sigma3=0,sigma4=loC,bt=1,plane=3,ncpu=nc).neo_f3kdb(grainY=grainY,grainC=grainC,mt=mt)
M = mt_merge(c,n,o,luma=true,Y=3,U=3,V=3)
S = c.bicubicresize(c.width/4,c.height/4).greyscale().levels(36*256,0.16,70*256,0,100*256).ex_invert().bicubicresize(c.width(),c.height())

I have not really looked too hard at it, but all of the "*256" stuff makes it seem 16 bit specific to me.


8 bit *1, # * [ 2 ^ (8-8) ] ::: * BitLShift( 1 , 8-8)
10 bit *4, # * [ 2 ^ (10-8) ] ::: * BitLShift( 1 , 10-8)
12 bit *16, # * [ 2 ^ (12-8) ] ::: * BitLShift( 1 , 12-8)
14 bit *64, # * [ 2 ^ (14-8) ] ::: * BitLShift( 1 , 14-8)
16 bit *256. # * [ 2 ^ (16-8) ] ::: * BitLShift( 1 , 16-8)

BPC = BitsPerComponent(Src)
MULT = BitLShift( 1 , BPC - 8 ) # Multiplier


Aaah yes I forgot this...

[guest]

Aaah yes I forgot this...

Hi again, anton,

So is this the reason why it's "fussy" ??

How can this be fixed ??

Cheers

[anton_foy]

Hi again, anton,

So is this the reason why it's "fussy" ??

How can this be fixed ??

Cheers

If you filter in 8bit then remove '*256' in the levels -lines in uband or '*1'. See StainlessS's post above for the different bitdepths.
What bitdepth do you filter in?
I will update this function soon to not be bitdepth specific.

[guest]

If you filter in 8bit then remove '*256' in the levels -lines in uband or '*1'. See StainlessS's post above for the different bitdepths.
What bitdepth do you filter in?
I will update this function soon to not be bitdepth specific.

Hi, I generally don't check the bit of the video I want to encode, but x264 is generally 8bit, but I mostly encoded to 10bit & x265.

I will check what "bit's" the video's are, that Uband doesn't like.

But your updated next build should be good to go :)

:thanks:

[StainlessS]

Just change to something like this, Untested.


# Uband - debanding script based on neo_f3kdb to affect only "flat / non detailed" areas
# Requirements: neo_f3kdb, fft3dfilter, ex_tools, masktools2, sysinfo

Function Uband(clip c, float "loC", int "grainY", int "grainC", bool "mt") {
loC = default( loC, 3.0 ) # smoothing strength of chroma in flat areas
# Neo_f3kdb parameters:
grainY = default( grainY, 23 )
grainC = default( grainC, 3 )
mt = default( mt, false )

mult= BitLShift(1,c.BitsPerComponent - 8)
nc = si_physicalcores()
O = c.bicubicresize(c.width/3,c.height/3).ex_edge(scale=3,"qprewitt").ex_expand().ex_invert().levels(0,0.1,212*mult,0,255*mult).bicubicresize(c.width(),c.height())
N = c.fft3dfilter(sigma=0,sigma2=0,sigma3=0,sigma4=loC,bt=1,plane=3,ncpu=nc).neo_f3kdb(grainY=grainY,grainC=grainC,mt=mt)
M = mt_merge(c,n,o,luma=true,Y=3,U=3,V=3)
S = c.bicubicresize(c.width/4,c.height/4).greyscale().levels(36*mult,0.16,70*mult,0,100*mult).ex_invert().bicubicresize(c.width(),c.height())
mt_merge(c,m,s,luma=true,Y=3,U=3,V=3)
return last
}

[anton_foy]

Very nice StainlessS many thanks I will test it later today!

[guest]

Very nice StainlessS many thanks I will test it later today!

Well, that's worked out pretty nice...

First of all for anton to put this little script together...

Me, for testing it, and finding it "fussy" with some types of video, and then commenting on it.

And for StainlessS, for chiming in, and "lending" his knowledge to figure out what was probably causing that issue :)