https://bugs.kde.org/show_bug.cgi?id=413125
--- Comment #9 from Estecka <stickman.c...@galacsys.com> --- Hi, I'm a long time user of Gimp, who just started trying out Krita 5.2.2. Krita's colour to alpha filter indeed works nothing like Gimp's, and no changing the threshold value can fix that: [Red to Yellow gradient, Orange removed](https://bugs.kde.org/attachment.cgi?id=170471) At a threshold of 43/255, Krita is both overly agressive on the red side of the gradient, andoverly lenient on the yellow side. The red side never reaches 100% opacity, while the yellow leaves a large strip that is 100% opaque. Both sides of the gradient still have traces of orange in them, when they should be pure Red or Yellow. > I don't know how is implemented in gimp 2.10 but it gives me the same results. As was mentionned, Gimp can be very finicky with its blending, so in order to get the expected result in Gimp you need to either: - With the Color to Alpha filter, switch the layer's composite space to "Perceptual", or switch the Blend mode to Legacy Normal. - Or, do not use the Color to Alpha filter, instead use the Bucket Fill tool with the non-legacy "Color Erase" blend mode. My expectation for this filter is that it should be the opposite operation to the "Behind" blend mode. Removing a colour and then backfilling with the same colour should result in a visually indistinguishable picture. (Safe for arithmetic imprecision.) I haven't checked the code for Gimp, but intuitively: if the Normal/Behind blend modes are linear interpolations between two colours, then Colour to Alpha/Colour Erase would be a form of linear *extrapolation*. Amongst the line of possible results, it would pick the first colour where any of the RGB channel has reached an extreme (either 0 or 255). -- You are receiving this mail because: You are watching all bug changes.
