Wavelet sharpening in GIMP

[cathouse42]

I found this plug-in for GIMP for wavelet sharpening.

http://registry.gimp.org/node/13438

Has anybody used this and if so, is it good?

[michael.h.f.wilkinson]

Haven't tried it, but gimp is limited to 8 bits per pixel processing, which is a limitation.

[cathouse42]

Haven't tried it, but gimp is limited to 8 bits per pixel processing, which is a limitation.

GEGL is 16 bit processing.

[ollypenrice]

Whenever I use sharpening I use it as a layer so that I can choose, minutely and meticulously, where I will and will not apply it. I think this is the most important thing to understand about sharpening. How you sharpen is less important in my view.

Olly

[D4N]

GIMP 2.9 is 16 bit (beta release) and I have used the wavelet sharpen feature in it successfully but as Olly says you have to be careful what you sharpen.

Stars can look awful if you over sharpen them and for planetary work you need to be careful to protect the edge of the disc.

Of course gimp is layer based so that's easy enough to do.

If you are doing planetary 8 bit might not be a problem anyway as often the camera is 8 bit and you don't need to stretch the image.

Sent from my iPad using Tapatalk

[michael.h.f.wilkinson]

Whenever I use sharpening I use it as a layer so that I can choose, minutely and meticulously, where I will and will not apply it. I think this is the most important thing to understand about sharpening. How you sharpen is less important in my view.

Olly

Many, if not most sharpening methods require such an approach, because they do not adapt their operation to the local image context. The same holds for smoothing, for that matter. You want to smooth very little in areas of high signal-to-noise (S/N) ratio, and much more in areas of low S/N. Many common methods are fixed.

My ideal approach is very different. I am developing adaptive sharpening and smoothing operators, which should be applicable to the entire image, and can be steered by the local image content, based on noise and object models. Still some way to go.