bruns added a comment.
Having had a look at https://code.woboq.org/qt5/qtbase/src/plugins/imageformats/jpeg/qjpeghandler.cpp.html#_M/HIGH_QUALITY_THRESHOLD, the jpeg handler actually know two different quality settings, `< 50` and `> 50`, the default being 75, i.e. high. High quality means floating point DCT and bilinear upsampling (Qt::SmoothTransformation), while low means IDCT and nearest neighbor upsampling. https://code.woboq.org/qt5/qtbase/src/plugins/imageformats/jpeg/qjpeghandler.cpp.html#337 https://code.woboq.org/qt5/qtbase/src/plugins/imageformats/jpeg/qjpeghandler.cpp.html#432 In either case, it uses libjpeg's capability to reconstruct smaller versions of the image, only decoding it partially, up to DC components only (i.e. 1/8 scale). So for a 24MPixel (6000x4000), libjpeg returns a 750x500 Pixel image, which is then downscaled by Qimage to 256x171 Pixel. Using SmoothTransformation for this size should be not to heavy. https://code.woboq.org/qt5/qtbase/src/plugins/imageformats/jpeg/qjpeghandler.cpp.html#297 I think best is to set the quality explicitly and mention in a comment what intended effect the setting has (so if the jpeghandler behavior changes, the code can be adjusted to have the wanted effect). REVISION DETAIL https://phabricator.kde.org/D26648 To: chroniceel, broulik, #frameworks, #vdg Cc: meven, volkov, cfeck, bruns, ngraham, kde-frameworks-devel, kfm-devel, pberestov, iasensio, fprice, LeGast00n, MrPepe, fbampaloukas, alexde, GB_2, Codezela, feverfew, michaelh, spoorun, navarromorales, firef, andrebarros, emmanuelp, mikesomov