Yes, that’s why I withdrew the post. I tried to step through an inverse, couldn’t get there.
1 Like
In the special case of a LUT (whether 1D or 3D HALDclut) that created different outputs for each input, inverting the LUT could be used to recreate the original, aside from rounding errors.
Inverting a LUT involves interpolation, because the LUT probably won’t have output values that correspond to every required coordinate in the inverse. This is easily solved for 1D LUTs but more difficult for 3D.
1 Like
The raw data is not affected by any camera effect or LUT.
Are you sure about that? Using CineStyle, all my raw photos are just as bland and flat as my jpgs (I have my camera save each photo to both formats, for various uses) - they’re all saved log. If I change my camera to another “picture style” (IE: one of the factory sets) my raw photos are yet again different.
This is kind of the reason for shooting in log (which is what the CineStyle LUT does). The image is stored with a flatter contrast allowing more room for image manipulation.
I think the problem is that LUT’s have only recently started becoming more common in still photography, even though they’ve been relatively common in a video workflow for a while. I can understand the frustration.
Well, I guess that makes sense. I didn’t realize that photography wasn’t quite up to speed on this like the video world is.
Picture style doesn’t affect your raw but it does determine the appearance of the corresponding JPG, and the embedded preview(s) and thumbnail(s). It is rare for camera settings to affect the actual raw image; rather, that info is stored in the metadata.
I think this was the solution.
- For the CineStyle LUT, use TC_Scurve.png.
- The adobe_2 and _3_strip.pngs are to simulate that old 70s Technicolor color film look you’re talking about, where they used rather large cameras that literally ran 2 or 3 reels of film at the same time to record the three base colors to separate strips (thus the name). We’ve come a long way!
- I have no idea what the TC_Dream.png LUT does.
By the way, the Technicolor CineStyle LUT is only available for Canon DSLRs apparently. Marvel Studios has another popular LUT with a slightly different look but I don’t know if it works on other cameras outside Canon.
Well, I don’t know what to say. On my Canon, my raw files look identical to my jpgs (excepting the level of quality, etc). The colors look exactly the same between the two.
Again, all the articles I read on the subject talk about such things as giving you more detail in the blacks, so you can crush them down to what you want since you can’t do the opposite of raising more detail out of them. The lights are easy; just don’t overexpose, but your blacks are at the mercy of your lights, thus, shoot it in log.
@Dennis_Alexander What file type are you talking about again? Maybe I got confused. If it is about stills, then my statement should hold true. If it is about video, then that is different. With video, most of the time, you wouldn’t be filming in raw, unless you have the money, gear and tech to support such a workflow. So, there are compromises. Log+LUT is one way to bridge the gap. That said, I am in no way an expert; I just read a lot 
.
Surprise surprise, you weren’t viewing the raw data in the raw photos, but only the embedded JPEGs. All raw photos contain embedded JPEGs regardless of camera setting. Some contain up to 3 embedded JPEGs of various sizes. Saving “RAW+JPEG” you’re actually saving raw + JPEG + JPEG ( + JPEG + JPEG), though the largest of the embedded JPEGs is not always of the same resolution as the standalone one.
No, raw is raw.
You’re welcome to upload a sample photo.
1 Like
How do you check that? It’s quite likely that image viewers don’t show the raw file at all but just the embedded JPEG preview.
Yes, that’s why I withdrew the post. I tried to step through an inverse,
couldn’t get there.
Ah, sorry, I was answering via mail and there I don’t see anything about
withdrawal.
1 Like
In regards to everyone saying that RAW files shouldn’t be messed with by the camera with a LUT, I understand what you’re saying and it makes sense. RAW is supposed to be completely uncompressed, and thus, large. And a JPG can be heavily compressed and if you set your camera to use a LUT, it will definitely alter the photo. But, here’s some screen shots.
This is a screenshot of a photo in Digital Photo Professional, Canon’s RAW photo editor. I’m pretty sure its not showing me a JPG.
I can’t even open a JPG in Canon’s DPP, so this is a view of the JPG in Windows Photo Viewer.
They both look identically log and flat to me.
Are you sure it’s showing you the plain raw data without applying the LUT you set in camera?
1 Like
DPP emulates the look of the in-camera JPGs and allows you to change the camera settings in-app after the shot. If you open the raw in RawTherappee and set the profile to neutral, you would see the raw as it is without the enhancements.
The only use I have for DPP is to see how the JPG would have looked by default. I normally set the camera to output as neutral a preview JPG as possible to get better histograms and metering, so the embedded images look flat and unappealing.
I haven’t used the modern iterations of DPP but I believe they have a few basic processing capabilities. However, given their limited features, I would refrain from using the app and stick with the more capable FOSS or commercial software.
1 Like
@Dennis_Alexander upload a raw file, not screenshots of a raw file.
It’s not surprising Canon’s raw editor is showing the raw file the same way a Canon camera processes the raw file.
1 Like
It isn’t showing the raw data. The raw data has a Bayer filter, so each pixel represents only one colour. So at the very least, the editor has de-bayered and white-balanced. As others say, it may also have applied any or all of the camera settings.
Here’s the original file.
IMG_2506.CR2 (23.4 MB)
When I go to load a profile in RawTherapee, all I get is a list of all the profiles I’ve created in the past. I don’t see a “Normal” profile.
I opened your raw with rawproc, using the dcraw equivalent “-O 0 -g 1.0”, that is, raw demosaiced data, linear gamma. I resized it for posting, saved it to a JPEG without an output profile:
Now, I reopened your raw, using the dcraw equivalent “-O 4 -g 1.8”, that is, raw demosaiced data with a camera profile (that’s what folks refer to as the ‘adobe_coeff’ table, the mother of all camera profiles table), converted to the dcraw ProPhoto colorspace, with the “recommended” 1.8 gamma, saved as a JPEG converted to sRGB with the corresponding attached profile:
Absolutely no processing other than the colorspace and gamma transforms, and the resize.
The first image is what your raw demosaiced data looks like. (Edit: Okay, I’m ignoring a few things, like white balance. If that were turned off, the image would have a green cast.) The second gives you an idea of what sort of things can go on in a programmer’s implementation of unadorned “raw conversion”
Had a few minutes before my granddaughter called me back to do more Snap-Circuits (electronics lab thing she just loves), so I thought I’d mess around…
2 Likes
Your CR2 has two embedded JPEG images, this is what they are:
- Thumbnail:
- Preview
Now let’s look at the image using the raw data:
- Here RawTherapee shows you the real raw data:
- And the most raw human-friendly form possible:
Likely the only effect applying the CineStyle effect has is setting this tag:
-TIFF-IFD0-ExifIFD-MakerNotes:CustomPictureStyleFileName=CineStyle
And most likely, enabling CineStyle has no influence on the raw data, though you would need to upload two photos of the same scene, one with CineStyle enabled and one without, to prove that.