Raw Therapee appears to be applying an image profile on RAW images, even with the profile set to neutral.


(Ingo Weyrich) #21

Wouldn’t it be better to report the issue here ?


(Ingo Weyrich) #22

I would like to mark this issue as ‘solved’ or to change the issue title…


(Morgan Hardwood) #23
  1. Step 1: Check what’s embedded in the raw file.
    exiftool -a -b -W %d%f_%t%-c.%s -preview:all _DSC6383.ARW
    
    In this case, two JPEG images are embedded in the raw file: a thumbnail and a large preview image.
    _DSC6383_ThumbnailImage
  2. Check the metadata. This raw file has more metadata than I’ve ever seen, so while I went through some of it and found no odd camera settings, I did not look through all of it.

While I admit that my initial reaction was that the RT preview looked not-neutral, I’m sure that’s just an illusion due to the high contrast scene. The image looks the same when viewed in RawTherapee, RawDigger, and other real raw processors.
What FastStone is showing you is not the raw image but the embedded preview JPEG.

Worth mentioning is that this Sony raw file has a raw curve which maps the 12-bit values to whatever libraw outputs (14-bit?).
Sony DSLR-A580.csv (58.3 KB)

$ gnuplot
set grid
set key autotitle columnhead
set datafile separator ','
plot 'SONY ILCE-6300.csv.txt' with lines


#24

One other difference vs other ARW test file I have : the ColorSpace tage is “sRGB” on all ARW files I have, excepted for your image where it is “Uncalibrated”, which might be the best suited option for uncommon color space.


#25

Oh, and RT also show the washed out image as thumbnail at first (using the embedded preview) before you apply the neutral image and get a natural looking thumbnail image.


#26

So one thing is sure : it’s not your lens, but rather camera settings.


(Andrew) #27

Is the purpose of this and the graph to convert the data from “s-log” (which @panomies mentioned), to linear?


(Ilias Giarimis) #28

This look of the in camera jpegs is very close to the look one gets by using slog2/3 profiles and not applying any correction … usually there is no mention of the slog curve used in jpeg’s embedded icc (TRC curve) so all editors just apply a wrong sRGB/adobeRGB gamma correction giving this low contrast image … If one applies the correct TRC correction (the inverse of the slog TRC) the jpeg picture will be very close to the contrasty neutral rendering of the RAW data.


(Ilias Giarimis) #29

The slog hypothesis gets stronger … by the 16/9 ratio of the jpeg image … :wink: