Processing RAW/DNG images from mobile phones

Hey @ggbutcher , it should be the ColorMatrix which corresponds to the D65 illuminant.

This camera uses ColorMatrix2 for StdA while RawTherapee expects ColorMatrix2 to be for daylight - see Derive correct ColorMatrix by checking CalibrationIlluminant · Issue #5126 · Beep6581/RawTherapee · GitHub

I observed no difference in the way RawTherapee renders the image nor in the way it calculates the temperature and tint when creating a camconst.json entry for this camera:

    {
        "make_model": [ "Realme RMX1851" ],
        "dcraw_matrix": [ 3516, -938, -625, -3281, 13125, 0, -1406, 4766, 5547 ] // ColorMatrix1 D55 from camera software: Realme/RMX1851EEA/RMX1851L1:9/PKQ1.190101.001/1566479375:user/release-keys
    },

@HimAgain don’t expect the image on your laptop screen to look like the image on your smartphone, i.e. don’t think that just because the image on the laptop looks different or worse than the image on the smartphone then your raw processing software is to blame. The screens on the two devices are most likely vastly different.

1 Like

@HimAgain It is important that the zoom is at 100% when you take the screenshots. When the zoom is less or greater than 100%, the quality goes down because the app has to interpolate pixels values. I will let someone else check your work and provide their pp3 and xmp for comparison.

Thank you, i know, but i don’t understand why i have at least 3 good profiles waiting for inclusion since months…

Lensfun, like many projects, is short on man power.

Ah, as opposed to the “media white point” for the Forward Matrix? I don’t see such in @HimAgain’s metadata, so I’m not sure how the ForwardMatrix can be used for his image…

Learn new things every day…

Is it not true that the shelf life of a smartphone is not more than couple of years? This is especially true when it comes to budget phones. On the other hand a digital camera is meant to be used for more than 10 years. So why would the lensfun team, already hard pressed for time, spend time on including profiles which will be used for such a short period?

1 Like

Well i don’t know, unfortunatly i don’t know anyone who can commit profiles, but i’m sure there is someone that has write access to the repo, maybe even here in the forum?

Anyway pictures took with a phone will remain even years… in case of my phone i first posted the vignetting profile in a request from november, 2017… and updating every time the xml is quite annoying.

For example i see a lot of lens db commits from sarunasb and junkyardsparkle . I have tried to mail Torsten Bronger to have write access but i got no reply.

And for example another one i made, the one for the Fuji 15-45mm, is the one for a well selling lens since it’s a kit one with really nice performances but really big uncorrected distortion, and being xtrans also ca correction is important since it isn’t autocorrected.

That makes sense to me that a color matrix in camconst.json shouldn’t be necessary if the DNG is reporting color matrix data. Although in this case, apparently handling “atypical” illuminants needs some work from the looks of that bug report?

@Entropy512 agreed that a matrix in camconst shouldn’t be necessary if the raw file already contains one. However, when a matrix is present in more than one possible source, currently it’s a black box as to which gets used, so RawTherapee could be improved by allowing the user to choose the matrix source, or at least to show the source, the same way you can sort-of choose the input profile (it could be improved as well).

Yeah, I need to do some digging (I have wound up far busier with other stuff than expected, I still need to get around to that second phase of scroll event investigations…), but I did notice that it seems like DNG metadata will take precedence over a DCP profile for example. (Or at least this seems to be the case with Sony A7M3 images that have come out of HDRMerge or manual DNG tagging, the latter of which I’ve done with siril output or HACK: Save aligned/merged Bayer image to TIFF file · Entropy512/hdr-plus@135273b · GitHub ) - Or it may be that in the process of tagging the DNG, I mangled some metadata in such a way as to cause the camera to be misidentified. Maybe I’ll get a chance to dig into this in a week or two.

1 Like

Everything has been uploaded, so keep on profiling lenses as soon or later they will be included!