Sign up now! · Forgot password?
RSS/Atom feed Twitter


This discussion is connected to the mailing list which is provided by the GIMP developers and not related to

1 of 1 message available
Toggle history

Please log in to manage your subscriptions.

Convolution DestinyLies 02 Jul 09:42
2020-07-02 09:42:34 UTC (about 1 month ago)
Send private message


Hi guys,

I hope this hasn't discussed before, but I couldn't find anything. My problem is that there seems to be a change from 2.8 to 2.10 in how Gimp calculates convolutions with arbitrary kernels (in 'Filter' --> 'Generic' --> 'Convolution Matrix') The easiest example I can come up with is a convolution matrix with only one enty in the middle >1, rest is 0, for example: [0 0 0
0 2 0
0 0 0]

The way convolution works this should only double every pixel intensity. If I use this convolution matrix in Gimp 2.8 on a grayscale image filled with pixels with itensity 50, the result is an image of pixels with itensity 100, as expected. However, in Gimp 2.10 the resulting pixels have a gray value of 71.. I deactivated 'normalize', no 'offset', 'dvisor' set to 1. I have read both documentations (for 2.8 and 2.10) but can't find anything about this different behaviour.. There seems to be an additional scaling, that I can't disable nor explain..

Can anyone of you help me? Thanks in advance and kind regards,