Magenta color in the shadows

I downloaded your raw file and was able to recreate the problem you are seeing in PL 3.3. I also opened it in ON1 Photo Raw 2020 and was unable to reproduce the issue there.

I was not able to recreate the problem on similar images taken with any of my Canon bodies. I have also not noticed it on numerous other images I have edited which were captured with other cameras and brands.

The attached image shows deep shadow detail retrieval from a Canon cr2 raw file taken with my 7D Mark II. I edited the image quick and dirty just as an example of raising shadow detail. There were no additional edits made to this image.

There is clearly an issue with your image and it is not related to the optics module because I tested it with and without it installed. Do you have the same problem with deep shadow recovery with all the images taken by this camera? DXO needs to review your image.


Deep shadows

Deep Shadow detail recovery

1 Like

I made a test with ACR: same magenta drift in the shadows as with PL.
Another test with ACDSee: the magenta drift seems to have disappeared …
Yet ACDSee is not a major reference in raw development. On the other hand, ACR is usually a reference. Oddly enough, correcting colored noise as much as possible with ACR … increases the magenta!
The particular point of Man’s photo is that it is underexposed, which the histogram shows (Note: this is not a criticism, just an observation).
This is from a photo from a 4x3 sensor, whose iso rise is relatively limited. To see the effect of correcting low light areas, I turned off SmartLighting, Selective Tone, ClearView and Contrast. It is then necessary to increase the exposure of +3. But if we also deactivate the vignetting correction (quite present at the edge of the image), it is rather +3.5 to +4 EV, which means that to have details in the shadows, we are no longer at 200 iso, but around 3200 iso! So with a lot of chroma noise to deal with. And it is perhaps at the level of the processing of the noise that lies the problem: ACDSee which leaves a lot of chromatic noise … leaves more green next to the magenta.
The platypus proposal works well, but I think we need to add Prime noise processing to it (probably looking for the best slider setting).

1 Like

However the problem also did not occur in ON1 when I opened up the shadows. Regardless that ACR had a similar issue with the magenta shift, .DXO should review this image.


Thanks for the analysis, i use zebra at 105% in my g9, but i think my g9 has a tendency to show zebra too ‘‘early’’. I’ll try to apply a permanent correction.
I have also slightly better result in Photoshop express (free version for android…) to recover the shadows…

it’s also a typical panasonic issue.
there shadows/near blacks have more magenta then it should be.
it’s off balance.
i had the same issue in night shots.
The darker spots got some magenta casting which i couldn’t repair. (G80)
i think your OOC jpegs are fine. (panasonic has long exposure correction which eliminates the redischnes.

DxOPL Staff did some readjusting of the Camera depending data.

Edit test with FastRawViewer and PL:

my example.

before g80 correction and after.

edit 2: see your rawfile exposure data:

it seems you could raise exposure to +.75 as in 3/4 of a stop.

(not that this help you for this image but you could use exposure bracketing for this kind of high dynamic scenery’s and find out in post which is best for development.)

1 Like

DxO has fixed color cast issues before (e.g., PhotoLab 2.2 and 2.0.1). They might be able to fix the magenta cast for the G9 also. The G9 issue (magenta cast when lifting shadows +3 EV or more) has been discussed on the Internet for over a year now, but maybe no one’s reported it to DxO as a bug yet. I’ve seen even worse performance reported for the Panasonic GX9 using ACR - hopefully that’s not a big problem with PhotoLab now. (It was discussed here as a PhotoLab issue a while back.)

I also suggest making sure your camera’s firmware is up-to-date.

1 Like

yes i have the same “issue” with my G80.
i turned it on a few months ago to see if it’s helping me and find it rather agressive.
It’s bound to the sRGB/ARGB colorspace i think and not to the rawfiles “colorspace” / luminancerange.
Same as the histogram feature in preview mode also sRGB (jpeg) mode and not raw.
i think all camera’s have jpeg sRGB histogram and overexposure/underexposure stripes preview and not raw based information preview.

Which is a nice feature to use even for raw is Idyn.
(it’s react on highdynamic scenery in auto mode by correction EV in -1/3 or -2/3 or -3/3. of a stop.)

1 Like

Thanks for the feedback
Yes it is up to date

I thought that I might give this file a try. I haven’t completed my trials yet but I have found a possible(partial?) solution. Not a cure by any means but better rendering at any rate. In the past I have had some problems with getting an orange hue when raising shadows by more than two stops. I found the cure for this is raising the shadow control in FP5 elite an equal or greater amount. This trick did not solve the magenta problem with this photo. I then tried the generic color rendering option “Neutral color/ realistic tonality (Gamma 2.2)”. This is the bottom option in the generic renderings tab. While it didn’t eliminate the magenta cast completely, it does make it at least tolerable.

Without my adjustment(color rendering):

Now with adjustment:


Final version after adding a little contrast, clearview plus, and smartlighting spotweighted on the problem areas in slight mode.

Even though this is a usable workaround, I agree with @mwsilvers that DXO staff should take a look at this problem. Perhaps just a small change in the appropriate camera/lens modules could solve the problem without having to employ this workaround.


I also sometimes see drifts in the shadows…
For me the normal solution is to use the curve of the tones, but what a mess with DxO VP ! It is necessary to correct the judged by going back under the histogram to see precisely the effect applied on the RGB values
This tool is a calamity in DxO PL level software
An adjustment (to be refined) for this photo: on the black point of the curves, set +7 of green and -4 of blue
I take the opportunity to renew here a demand for evolution of the tone curve to have at least at the level of the average of the software of processing of the RAW


Despite PL’s magenta drift behaviour, which I also encounter from time to time on a Lumix G9, Man’s original picture is nice example where ETTR could have avoided this issue right from the start. It seems that there was plenty of room left for a positive EV correction.

1 Like

Hi, i do ETTR, based of zebras set to 105%. But i realized with a set of pictures including this one that the zebras, even at 105% setting, are not reflective of the raw histogram. Same for blinkies in replay.
Fast raw gives me a margin of 1EV/1.33 before having burnt highlights.
i therefore now will bracket in this kind of high dynamic scene.
Nevertheless, results to recover shadows without the magenta cast are better in other softwares.

The magenta drift here is not only linked to PL since we have exactly the same problem with ACR (Ligthroom). But it would be nice if PL found a solution …

1 Like

HELLO the sensitometric curves are not linear sooo you have to control with the U point ! so it is normal ,


Hello Payen,

Welcom to the forum.
would you care to elaborate? (give some more information?)
I had to look up “senitometric” and i assume the controlpoint you talk about is for lifting and colorcorrection purposes?

I am in Belgium
i am retraited about 40 years in film industry like Dop .
I have a point of view about digital technology, it is not magical ! , the response of a sensor like ours Canon or Nikon or PeNTax or etc characteristics … it is not " rectilinear " in term of Sensitometry ( the science measuring the response of illumination and so gamma of the sensor etc…) So you have always some distortion in color science .
I have notice a deviation of colorimetry, in resume in extreme contrast like your exemple 200 iso you are about a Dynamic of more than 13 stops , the cMos can t take this extreme level … even at 100 iso
A question : But why do you want informations in this ultra black and noisy zone !!! ?? No interest for me indeed you ll want to much for any reasons ??? the subject is not the black ??
Only a bracketing in over exposure and under exposure with HDR or similar Soft wares ( only applied with subjects that are not moving ) can take the “informations” of burning areas and the blacks in extreme lighting contrast ! .
The other way to do this kind of exercice is to put lights in dark areas 1 with Large White tissues in reflections all the DOPin films are putting lights in the shadows to reduce contrast ,you ll never have the place to put these !
2/ HMI 18 kw ( kind of high pressure metalloid iodure projectors to compensate the hard sun ) you need a elect generator it is not for a standard journey !? you have to put it out of the frame or erase it digitally
3/ more simple for photographers the 500w flashes with light box or diffusors !
In my opinion only the new sensor BSI are capable of this performance with one shot in E I ( exposure index of 64 iso )
Like the Z7 ou Lumix S1H can make the job at 32 iso or 64 …
But why ? the subject is the castle and the sky …the are nothing in your black …perhaps a ghost? !
With my best
Payen Patrice


I understood your goal and explanation.

To be clear this wasn’t my shot (castle) but most people like me also have a view of the scene which is HDR-erd in the Brain. we scan highlight and shadow wile the pupil adapting to the reflected light so we think that’s the “real scene’s dynamic range” and we like to replicate this in post as the mental picture we have. (rawdevelopers allow us to stretch Dynamic Range a lot at the cost of errors and artefacts.)
The only way to do that correct is exposure bracketing and stack like our brain does.
My “issue” was and is for a little is that nightshots like mine above of the habor has some purple/magenta cast also. Panasonic has lowlight/long exposuretime corrections build in to remove this in the OOC-jpeg but the rawfile doesn’t have the correction build in so it have a haze of redischness in some darker area’s.
for my G80 rw2 DxO did re-callibrated the profile so it has less redisch glow/cast/haze.
That was my initial point: it’s a panasonic issue in the rw2.
BSI and 64iso as base iso? in my m43? wowwh that would be stunning! :heart_eyes:

1 Like


Just made a quick comparison with some softwares i have access to : ACR, Affinity, Dxo on PC. Just boosting shadows (shadows + blacks in ACR, pushing blacks in Dxo is …ugly) : Dxo introduces much chrominance noise in the shadows in the exemple picture with the castle (which is just an exemple) , ant that, amplifies the magenta cast.
Of course can be worked through different methods in Dxo, more or less efficient and quick (TSL and/or curves in local adjustment would help). But on the whole i have the feeling that, regarding shadows recovery, with my panasonic G9, Dxo is not the best software.

just one reminder to be comparing in an exact manner the best way is do a export on all applications and compare the jpeg.
DxO does a lot in Prime and export rendering and 75% zoom isn’t always show the full picture’s quality.
So if i doubt a certain denoise setting or other things as sharpening or microcontrast things (clearview) or faces and fine contrast i just export with suffix “test contrast” or “test noise”

i am in this case interested in a jpg export of all and side by side in faststone image viewer of your findings.
(just every application a 5min development time which is about the random time you use for not really important images which has some difficulty’s.)
Not to bash a application but to learn something.
(my old Silkypix doesn’t take G9 rw2’s)