Hi there,

Following a cool tutorial from Patrick David (see : http://blog.patdavid.net/2013/02/calvin ... -gimp.html), I've added a new filter 'Colors / Freaky details' in the G'MIC plug-in for GIMP (added it to G'MIC online as well).
This effect allows to enhance details in your images, it looks like this :


For curious people, here is the G'MIC code I wrote to achieve this effect. This is a quite straightforward translation of the tutorial in the G'MIC language : http://pastebin.com/jTGKWVdr

I hope this will be useful, if you have any examples of use, please let us know !

The result shown sort of looks like a cleaner version of Adaptive Contrast (ACE) filter David. Cool stuff. Had a problem visiting Patdavid's link that you shared; might be because I'm still using IE8. lol

Darnit! I was going to use this process to learn G'MIC programming!

Well, maybe I'll go record a video tutorial of the process instead.

Hi Ronounours,

I experimented and compared the results, changing Selective Gaussian Blur, Smooth [bilateral] and Freaky details.
Then I will try to use this image in PS and compare the results.

Selective Gaussian Blur


Smooth [bilateral]


Freaky details.


Smooth [bilateral] + Walvet Sharp


Particularly liked the most when I added the walvelet sharpen.

The result obtained with the Freaky details, was somewhat dimmed when compared with other results. Even compared to the original result (http://blog.patdavid.net)

Ronounours and patdavid, Thank you for rescuing this technique of patterns PS and bring to Gimp!

Need to check that indeed. Thanks for your testing.

This is what I get running G'MIC Freaky Details (2, 1, 0):


It's pretty close to the original method from the tutorial. Not sure why yours is so dim?

Yes, Pat I get the same image a yours, with the default settings of 'Colors / Freaky details' (with G'MIC 1.5.5.0 beta).
Maybe Akros is running an older version of G'MIC where some bugs appear during the process ? Could you tell us what version of G'MIC do you use ?

I should note, though, that I do get different results if I run it on a larger image.

The spatial and value variance will change with scale of the image being applied. Perhaps set that as an input option on the filter instead of just Amplitude and Iteration?

For instance, I'll use very different variances on the bilateral smoothing step depending on the input size of the image...

Yes, good idea. I'll add some new options to the filter tonight.

I used default settings too.
version 1.5.3.0

@pat : I've added a new parameter 'Scale' for setting the sigma of the bilateral filtering.
@Akros : Maybe that's the reason. Could you try to update to 1.5.4.0 to see if that fixes the problem ? Thanks.

I've done these attempts:

Pat David tut without G'mic:
Duplictate → Invert → Gaussian selective (blur ratio:9) and set layer in soft light → Visible → Apply Dodge&Burn script to Visible → Visible#2, set in overlay mode.

Pat David tut Anisotropic Smooth instead Bilateral:
Values: Amplitude 40; Sharpness 1,0; Gradient smoothness 1,0

Pat David tut Freaky details:
Amplitude: 2, Iterations: 1 , Channels: Luminance and a touch of gimp's unsharp mask (ratio: 30, amount: 0,70, treshold: 0)

To be fair, I'm on 1.5.2.4 myself...

Eureka!

I had previously obtained image via right click >copy image then with gimp: file - create - from clipboard

This way of obtaining the image before me had already caused problems with transparency (commented at the time here in the forum)

So, I changed the way to get the picture, right click - save the image as!
And using exactly the same steps, the result is different! Thus, the result is not as dark as it is obtained via image copy.

I need to stop using this form of obtaining images. But the intention was only to edit and check out the tutorial, I did not need to save it on my pc.

I only search and found this thread after writing having written script for "Freaky Details" following Pat David steps.
https://gimplearn.net/viewtopic.php?t=483