I'm not very surprised to see that you can get better results than the default color profiles. The out of the box colors for those Sony cameras are not very natural. Blue colors are often looking bad on a lot of Sony cameras (Nex, A6000, A6300, A7...). I'm almost never using the Sony Alpha curve. Best solutions to me are either to go for Neutral base curve, or a bit like your idea, Linear 709 as the input profile, with a lot of added saturation. So I'm quite interested by your work : ) To extract manually the profiles seems to be a bit complicated... So I will keep reading this thread !
François On Fri, May 4, 2018 at 9:11 PM Sarge Borsch <sausagefacto...@gmail.com> wrote: > So, I tried to edit an ICC file in a hex editor to put the values from > that website, and, just as expected, got nonsensical results. > > After that I tried another idea to snatch better color profiles — I > searched the web for Adobe Camera Raw package, extracted the profiles from > it (they are in .dcp format), and figured that it's possible to convert > them to ICC by dcamprof. > They seem to work very well — better than the currently built-in input > profiles for sony a5100 in darktable. > Now what do you think about the copyright status of these converted ICC > profiles? Can they legally be distributed with darktable, or should I keep > them only for myself? > They are a lot smaller than the source .dcc files, probably because they > don't keep nothing valuable except the color matrices. So are 3x3 numeric > matrices copyrightable? > > If you think these profiles can be officially added to darktable, I may > fix the name tags and submit a pull request. > > > On 4 May 2018, at 17:14, Sarge Borsch <sausagefacto...@gmail.com> wrote: > > > > Hi. > > I see that there are measured color responses at > https://www.dxomark.com/Cameras/Sony/A5100---Measurements > > Hence the question: is it a good idea to try to take the built-in > profile and replace the primaries with these measured values, in order to > get closer to the in-camera JPEG color rendering? (or are they already > used?) > > > > I'm asking that because I've noticed that none of the built-in input > color profiles for sony a5100 allows me to get close to the in-camera JPEG > colors. > > The 2 of them which are the closest to the correct rendering (that is, > matching camera JPEG, which is quite good when judging by eye) are the > "standard color matrix" and "linear Rec2020 RGB". > > Both of them wildly differ from the in-camera JPEG in deep blue colors: > "standard color matrix" causes them to be clipped and to look really > unnatural, and "linear Rec2020 RGB" looks more or less natural, but the hue > is obviously different (blue gets moved to cyan). Hence I started to wonder > how easy is it to get a better color profile. > > > > I know that ideally this should be done with a color chart, but I don't > have one and don't have spare money for it at the moment. > > > > Also I can share a shot of the example object (Raw + JPEG) which has > such problematic color if anyone wants to test it, too. > > ___________________________________________________________________________ > darktable developer mailing list > to unsubscribe send a mail to > darktable-dev+unsubscr...@lists.darktable.org > > ___________________________________________________________________________ darktable developer mailing list to unsubscribe send a mail to darktable-dev+unsubscr...@lists.darktable.org
