On Tue, Jan 17, 2017 at 6:46 PM, Jan Buchmann wrote:

I saw in GIT that you are working on the layer modes (linear vs gamma etc) currently. For some key adjustments that I do, I use the vivid light layer mode, which is only available via calling GMIC at the moment. Would it be hard to implement it also in core GIMP (GMIC seems to have an implementation)?

https://bugzilla.gnome.org/show_bug.cgi?id=775857

Thomas would have to update that patch given all recent changes, though.

Alex

On Sun, Jan 22, 2017 at 3:29 PM, Alexandre Prokoudine wrote:

On Tue, Jan 17, 2017 at 6:46 PM, Jan Buchmann wrote:

I saw in GIT that you are working on the layer modes (linear vs gamma etc) currently. For some key adjustments that I do, I use the vivid light layer mode, which is only available via calling GMIC at the moment. Would it be hard to implement it also in core GIMP (GMIC seems to have an implementation)?

https://bugzilla.gnome.org/show_bug.cgi?id=775857

Thomas would have to update that patch given all recent changes, though.

And another update: pippin independently added vivid light and more blending ops to Git master:

https://git.gnome.org/browse/gimp/commit/?id=c9ad09df6195e6ed59f3ece17b055eaed5891e22

They haven't been exposed in UI yet. That's on TODO list.

Alex

Hi all!

While you guys are at it, I encourage you to try out this formula:

E = (2 * I ) - M

I call it the "Opposite" mode. I derived it from how the unsharp mask does the sharpening - first run a grain extract and then do a grain merge.

It is not useful in an awful lot of places. But when you want to de-convolute in a strange way, and you have the procedure or a plugin which does only the convolution, "Opposite" layer mode can help you get there quickly!

It also has the "opposite" effect on most color effects as well!.

Cheers!

Laxminarayan Kamath Ammembal
http://lankerisms.blogspot.com
(+91) 9945036093

On Mon, Jan 23, 2017 at 12:58 PM, Laxminarayan Kamath wrote:

Hi all!

While you guys are at it, I encourage you to try out this formula:

E = (2 * I ) - M

I call it the "Opposite" mode. I derived it from how the unsharp mask does the sharpening - first run a grain extract and then do a grain merge.

It is not useful in an awful lot of places. But when you want to de-convolute in a strange way, and you have the procedure or a plugin which does only the convolution, "Opposite" layer mode can help you get there quickly!

This is similar to something else I have experimented with, permitting opacity values below 0% and above 100% for normal mode, this permits implementing unsharp mask with a blur layer - as well as increasing saturation by blending a color image with it's desaturated variant.

/pippin

On second reading, the previous description is not that clear on how to use it. (unless you have read the unsharp mask script)

You have an image with a slight radial or zoom blur or motion blur (the photographer tried to be creative) . You duplicate the layer.
You measure and run the same exact blur again on the duplicated layer! Say the radial blur is about 10 degrees. You run the radial blur at 10 degrees. In the same direction as the original!

Then you switch the layer mode to "Opposite". Voila! if you measured right, the zoom/motion blur is reduced significantly!

For color effects it works as well, like for "add punch" from FxFoundary color effects,it removes the punch, etc.

For other kinds of effects, it gives various interesting results - for edge detect - instant painting/cartoon like effect :)

On 1/23/17, Øyvind Kolås wrote:

This is similar to something else I have experimented with, permitting opacity values below 0% and above 100% for normal mode, this permits implementing unsharp mask with a blur layer - as well as increasing saturation by blending a color image with it's desaturated variant.

/pippin

Nice! looking forward to it :)

Laxminarayan Kamath Ammembal
http://lankerisms.blogspot.com
(+91) 9945036093

On 01/23/2017 06:58 AM, Laxminarayan Kamath wrote:

Hi all!

While you guys are at it, I encourage you to try out this formula:

E = (2 * I ) - M

Note that the recent work in master removed the channel-value clamping that used to happen during blending (yay! :), and since each layer appears only once in this formula, you can implement this mode using the existing ones, with the help of some ancillary layers, but without having to duplicate the "content" layers (I and M).

-- Ell