Re: [Gimp-developer] Gimp from git Gegl Gaussian Blur and Unsharp Mask useability issues

[Teo Mazars <mazarst ensimag grenoble-inp fr>]

Hi,

I have seen the comparison between the old (from gimp 2.8) and the new gaussian blur. I did not have a look 
at 2.8's implementation yet, but one of the main difference, is that the new implementation works in scRGB 
instead of sRGB in GIMP 2.8.

I don't know if you were aware about that. I personally would still let the possibility to make it work in 
both sRGB and scRGB, depending on user's need. Both behavior are useful. But it's still under discussion.

I completely forgot that until now, but Massimo worked on Gaussian-Blur some times ago, and his work is now 
in operations/workshop (you have to pass --enable-workshop to autogen to build them). This implementation 
should replace the current implementation at some point, but the behavior is equivalent to the current 
implementation. I mean all the work you did still applies on the new implementation. The main thing that 
prevents the replacement is that the case with std-dev < 0.5 is not implemented.

Among all improvements: Waaay faster, even with very large std-dev, it allows setting how edges are handled 
and finally it fixes the "vertical streaks" bug you were seeing with large std-dev.

I don't answer to the rest because I don't know or I did not though about it yet. Though, we will need to 
find the equivalent conversion between the 2.8 gblur and GEGL, because 2.10 keeps api compatibility with 2.8 
and that includes that all filters should be have the same given the same set of input parameter from the api 
point of view (that is, not necesserly in the ui).

Regards,

Téo