Reported gamma errors when scaling

[speculatrix]

I was wondering whether Irfanview did its scaling correctly?

http://www.4p8.com/eric.brasseur/gamma.html

http://www.doxdesk.com/updates/2008.html#u20080415-resize

[Edogy]

Been using IV for years, wonderful software. I saw the page speculatrix linked above on reddit today & came here to inquire about the same thing. I took that first image in speculatrix's first link and resized it to 50% in IV 4.25 and got:

[MItaly]

I don't have IV at hand now, but which resize mode did you use?

[Edogy]

Correction: The solid gray image I linked above is the result of a IV resample operation. If I resize instead of resample I get this:

Which suggests the way IV handles gamma may be incorrect, as stated by the original links. In both cases the image I am resize/resampling is this one, specifically crafted to highlight gamma errors in resize/resample algorithms:

[jbw]

Not only scaling, but perhaps every filter with averaging colors of neighbour pixels is affected with this bug. Complementary colors makes the most dramatic effect. Let's take a picture:

then apply "zoom blur":

whereas the right effect should look like:

Both are done with IrfanView, but in the last case I've done what IV "forget" to do: flatten gamma curve before applying filter and restore it after.

[jbw]

This simple trick works, of course, with scaling of DL picture, too: apply gamma 0.45, scale image to 50%, apply gamma 2.2, and... tada!

It would be very nice if IV could make it itself on behalf of us ;-)

[eL_PuSHeR]

Do you know if this bug apply only to downscaling or also to upscaling?