Camera Support #10565
Please add support for this camera. I see that dcraw 9.26 supports it (including color matrix), and I'll help out with base/tone curves and WB presets as soon as the one I just ordered arrives.
#2 Updated by Ralf Brown over 3 years ago
- Status changed from Incomplete to Patch attached
- % Done changed from 20 to 70
Adding the following entry to share/rawspeed/cameras.xml gives a not-outlandish rendering of the image, though of course the colors are off (way under-saturated, mostly) without the color matrix.
<Camera make="OLYMPUS CORPORATION" model="TG-4"> <CFA width="2" height="2"> <Color x="0" y="0">GREEN</Color> <Color x="1" y="0">RED</Color> <Color x="0" y="1">BLUE</Color> <Color x="1" y="1">GREEN</Color> </CFA> <Crop x="0" y="0" width="0" height="0"/> <Sensor black="250" white="4095"/> </Camera>
Unfortunately, two attempts to upload the ORF failed (at what looks like the 4MB mark), so I've uploaded it to my photo-sharing account: http://imageevent.com/ralf/olympustg4
#4 Updated by Ralf Brown over 3 years ago
Pull request updated with noise profiles. I had trouble generating base/tone curves -- the camera is applying lens distortion correction to JPEG output, which is confusing dt-curve-tool since the images don't quite match up.... However, from the flawed curves I generated and visually matching Darktable's rendering against camera JPEGs, I can say that the base curve differs substantially from the "olympus-like" preset (darker shadows and less aggressive brightening of highlights) and even more so from the alternate preset, which really crushes highlights.
#5 Updated by Pascal de Bruijn over 3 years ago
I've merged the pull request, with some minor added-finetuning:
@basecurve, did you try to see if any of the other brand curves do any better?
#6 Updated by Pascal de Bruijn over 3 years ago
Taking another look, it would seem asif the image needs a tad of cropping as well:
The sample you provided makes it really difficult to get pixel precise value. Presumably you can test if I got it right on some of your images? Please do try with -12 as well.
Also, I'd love to see a ISO100 and ISOmax sample as well.
#7 Updated by Ralf Brown over 3 years ago
The OoC JPEGs are 4608x3456, so it looks like the camera crops 16 pixels from each side and 8 from top and bottom. I hadn't noticed any problems at the edges, though, even without specifying cropping in cameras.xml.
I've added the OoC JPEG for my initial sample image to my ImageEvent album, as well as JPG+ORF for shots taken at ISO 100 and 6400 as part of noise profiling.
Please do try with -12 as well.
Do you mean the crop value in cameras.xml? As I said, I didn't notice any issues at the edges. If you want to exactly match the camera's JPEG dimensions, the values to use would be x=16,y=8 and the corresponding 16/8-pixel crop on the right (are the width and height values in cameras.xml relative to the uncropped image or the specified origin?).
#8 Updated by Pascal de Bruijn over 3 years ago
Matching the JPEG typically isn't the goal for any open source RAW converter. Typically only obviously bad bits are cut off.
Would you mind supplying some more ORF samples so I can take a look myself, preferably a whole set of (outdoor) shots per ISO setting, so we can check blacklevels as well.
#10 Updated by Ralf Brown about 3 years ago
- File lensfun-oly-tg4.xml added
Hmm, it seems that my previous message never got posted -- are you still looking for the outdoor ISO series, or is the blacklevel info from the ExIF that you're currently using sufficient?
And just to keep all the info for the TG-4 in one place, the lensfun database entry I generated today.
#11 Updated by Pedro Côrte-Real about 3 years ago
- Status changed from Patch attached to Fixed
- % Done changed from 70 to 100
The support should be mostly done now. If you could submit one or more samples to rawsamples.ch that would be appreciated. We use that collection to do regression testing and make sure everything keeps on working in the future.