Conditional filter

[Terka]

Have a big grain that differs from background only a little.
Is it possible to create such condition:
if all pixels in radius are similar (differences below threshold) do averaging/blurring.

[cretindesalpes]

Not sure if it helps but you could try this:

# Parameters: rad, thr

# if all pixels in radius are
range = Dither_build_gf3_range_mask (rad)

# similar (differences below threshold)
mask = range.mt_binarize (threshold=thr, mode="upper")

# do averaging/blurring.
bl = RemoveGrain (12).RemoveGrain (20)
mt_merge (last, bl, mask, luma=true)

[Terka]

cretindesalpes, thank you for help.
But are you sure the Dither_build_gf3_range_mask does exactly this:
eg radius 1pixel = 3x3 area , values
100 102 105
100 103 105
100 105 105
threshold 10. Pixel should be averaged.
Could radius "cross shape" be used?
abc
def
ghi
be taken?

[Didée]

if all pixels in radius are similar (differences below threshold) do averaging/blurring.
Do you think that ^this^ is really so blatantly different from the much more simple:
"Do averaging/blurring with all pixels within [radius] that are similar by [threshold]"

For the latter, the simple answer is DEEN.
(And has been for 10 years, to date.)

[Guest]

For the latter, the simple answer is DEEN.
(And has been for 10 years, to date.) Indeed. And Smart Smoother was doing it as a VirtualDub filter for 13 years.

[Terka]

Thank you.
In fact im trying to denoise this:
http://s23.postimg.org/z1hhxf2jv/nikond800_iso.png
Or if necessary denoise, resize down and denoise, because the grain is too big - eg purple grain

[Guest]

In fact im trying to denoise this:
http://s23.postimg.org/z1hhxf2jv/nikond800_iso.png There are 4 photos there. Which is the original you are trying to denoise?

[Terka]

the ISO 6400 and 12800

[Guest]

And how do we download the original PNGs?

[Terka]

I have tried on this small sample, cropping the picture i uploaded.