RW2 files with important magenta color shift

Discussion in 'Panasonic Cameras' started by Zeus1, Jan 11, 2016.

  1. P1010038_IJFR. Bought a Pany GM5 as "point and shoot" to supplement my Oly OMD EM5.
    Unfortunately all the RW2 raw files show an impressive color cast (see the included image). Now, this is not a major issue (I always shoot in raw, so color correction is only an extra step in post-processing). On the other hand, I really like the colors of Olympus raw images; color correction is practically never necessary.
    Searching the internet revealed different instances of magenta or green color shifts with Panasonic bodies, but never with a GM5.
    What are Your ideas?
  2. eteless

    eteless Mu-43 All-Pro

    Jun 20, 2014
    I've seen a slight blue shift with a camera that had it's sensor cleaned with windex which left a blue residue on the surface, however it was uneven over the frame (drying marks etc).

    No clue why it's so magenta, gl!
  3. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    Real Name:
    I'm seeing very little of that on my tablet. there is little information I can get from the image, what were your shooting parameters (like colour balance setting? was it set to tungsten?)

    just for shits-n-giggles try running it through draw (free to download) and see what its defaults are (or load a raw file for me to have a quick crack at it)
  4. hazwing

    hazwing Mu-43 All-Pro

    Nov 25, 2012
    Just to rule out... your not using a polariser or other filter are you?
    How are you viewing your RAW files, is it through lightroom? Do you get the same colour cast in jpg, when shooting RAW+jpg?

    I've previously noted a green colour cast to my GM5 RAW files in lightroom. I think I've just become adjusted to it now.
  5. @pellicle@pellicle: RW2 files contain only minimal EXIF info. Posted an image in my dropbox, please feel free to d/l and examine.
    Dropbox - P1010082.RW2

    @hazwing@hazwing: Using an UV filter. But have taken images with and without this filter, always the same magenta cast. Viewing raw images in Photoshop and DXO optics Pro, latest versions. Yes, same color cast in raw + jpg images.
  6. b_rubenstein

    b_rubenstein Mu-43 All-Pro

    Mar 20, 2012
    Melbourne, FL
    I bought one of these when they were on sale for $50 last month. Very easy to create profiles for a camera. Supports DNG profiles used by Adobe products; don't know about other RAW converters.
  7. wjiang

    wjiang Mu-43 Hall of Famer

    I noticed something similar between my GM1/E-M5 previously and GM5/E-M1 now. It might just be how Panasonic and Olympus are tuned differently by default within your converter.

    With ACR I've just set the colour profile for the Panasonics to always apply a green/magenta shift to compensate, so that they come out closer to the Olympus files with the same WB setting.

    Incidentally, I've also dialled back the reds on the default colour profile applied to the Olympus because I always found them oversaturated compared to other colours.
  8. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    Real Name:

    firstly I misread your original post and theought you were talking about a G5 ... your issue is something I have found in the analysis of other recent pany camera RAW files ... so its not surprising to see that here. I've put a blog post over here: in my view ...: noise about sensor noise discussing that. Essentially I think Panasonic are tuning the shitter out of their "image data space" to minimize noise (and perhaps fudge a few things ...)

    So, firstly, the cunning thing with RAW files is that they have the JPG embedded in them, which is used by the camera to show you the camer settings in "play" and also zoom in ... its a half full size image. dcraw extracts that and so here is the EXIF
    Filename - p1010082.jpg
    Make - Panasonic
    Model - DMC-GM5
    Orientation - Top left
    XResolution - 180
    YResolution - 180
    ResolutionUnit - Inch
    Software - Ver.1.1
    DateTime - 2016:01:01 13:12:45
    YCbCrPositioning - Co-Sited
    ExifOffset - 634
    ExposureTime - 1/60 seconds
    FNumber - 4.20
    ExposureProgram - Aperture priority
    ISOSpeedRatings - 1250
    ExifVersion - 0230
    DateTimeOriginal - 2016:01:01 13:12:45
    DateTimeDigitized - 2016:01:01 13:12:45
    ComponentsConfiguration - YCbCr
    CompressedBitsPerPixel - 2 (bits/pixel)
    ExposureBiasValue - 0.66
    MaxApertureValue - F 4.17
    MeteringMode - Multi-segment
    LightSource - Auto
    Flash - Flash not fired, compulsory flash mode
    FocalLength - 18.00 mm
    SubsecTime - 042
    SubsecTimeOriginal - 042
    SubsecTimeDigitized - 042
    FlashPixVersion - 0100
    ColorSpace - sRGB
    ExifImageWidth - 1920
    ExifImageHeight - 1440
    InteroperabilityOffset - 13652
    SensingMethod - One-chip color area sensor
    FileSource - DSC - Digital still camera
    SceneType - A directly photographed image
    CustomRendered - Custom process
    ExposureMode - Auto
    White Balance - Auto
    DigitalZoomRatio - 0.00 x
    FocalLengthIn35mmFilm - 36 mm
    SceneCaptureType - Standard
    GainControl - High gain up
    Contrast - Normal
    Saturation - Normal
    Sharpness - Normal
    Maker Note (Vendor): - 
    Image Quality - Raw
    Focus Mode - 
    AF Mode - 
    Color Mode - Normal
    Faces Detected - 0
    Face Info - 
    Country - 
    Region - 
    City - 
    Landmark - 
    County - 
    Thumbnail: - 
    Compression - 6 (JPG)
    Orientation - Top left
    XResolution - 180
    YResolution - 180
    ResolutionUnit - Inch
    JpegIFOffset - 14836
    JpegIFByteCount - 10932
    YCbCrPositioning - Co-Sited
    then the image from that:

    now I whacked it into dcraw and got interesting issues visible when I did a straight demosiac (convert from RAW to a valid 3 colour per pixel image)

    Looking at that in Photoshop shows the movement of R G and B colours to strange places (in the past they were balanced in the demosiac relative to each other without requiring much "moving" {scaling})


    Unlike what I found on the GX7 images these seem more "central" ... I suspect they are moved over to avoid floor noise in the shadows ... perhaps some non-linear algorithm assists that in scaling.

    I moved the dark point (without clipping it) then did an auto colour in photoshop to get this:


    which supports the ideas already suggested that colour profile of the camera will give good quick results (if happen to have profiles and you don't want to use the camera JPG).

    Myself I use RAW exclusively and pull out the JPGs with dcraw -e *.rw2 to give me a smaller JPG which suits 99% of what I want. Then if there is something I really like and I want to print it I'll fart around with other tools (like dcraw or rawtherapee) and get what I want. dcraw is free to download, for all platforms (Mac, PC, Ubuntu (or any linux you like) and if you do a -T -4 extract you'll extract a linear 16 bit tiff that you can throw into (almost) any version of any image editor (and you won't need to muck with dng or whatever to get (your older version of photoshop) to access the RAW files. You can tell I'm an engineer can't you ;-)

    Hope that helped...

    Best Wishes
    Last edited: Jan 11, 2016
  9. Turbofrog

    Turbofrog Mu-43 Hall of Famer

    Mar 21, 2014
    Interesting. Never noticed this with the GX1 or GX7. If this is a common issue, that would definitely temper my enthusiasm for the GM-series. Seems like it has to be a camera defect...
  10. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    Real Name:
    myself I don't view this as a defect, just a departure from a tradition of mapping the data down (from each or R G G and B sensors in the bayer array) to best suit the needs of sensor to give the best image possible with the sensor technology at hand.
    From there its just signal processing to get it where it should be ... for what its worth in a 16bit linear demosiac there is less wasted than appears to be wasted over in the LHS to noise because of this.

    And I forgot to post the EXIF data that I pulled from the extracted JPG, so I'll amend the above post to reflect that
    Last edited: Jan 11, 2016
  11. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    Real Name:
    call me Arg
    This is probably all about your raw processor. Try other processors and see if you get the same.

    Shoot raw+JPEG and see if they both have the same tint.

    Shoot a grey card on Auto WB, raw+JPEG, and see what you get.
    • Agree Agree x 1
  12. wjiang

    wjiang Mu-43 Hall of Famer

    Optimising per channel dynamic range seems like a valid and worthy reason for the difference.

    I'd actually blame the RAW converter, as the RAW conversion is entirely up to the software. Custom profiles sets like Huelight will equalise the differences between bodies out. The OOC JPEGs from my GM5 are great, problem is that the RAWs don't look like that when first loaded into ACR with the available Adobe Standard profile. Olympus files sucked too until Adobe added Camera simulation profiles (e.g. Camera Natural) for the Olympus bodies. Perhaps we should somehow get Adobe to make some Panasonic profiles too.
  13. T N Args

    T N Args Agent Photocateur

    Dec 3, 2013
    Adelaide, Australia
    Real Name:
    call me Arg
    For my Panasonics I created LR camera profiles using a Colorchecker target.

    Olympus too, actually. :2thumbs:
  14. jyc860923

    jyc860923 Mu-43 Hall of Famer

    Feb 28, 2012
    Shenyang, China
    Real Name:
    back in 2012 when I was shooting GF3 and using LR4 to develop the rw2 files, I had to move the red (the first one in hue) slider to +25 to get the right colour, then the same with G5's rw2, but probably later LR versions changed and I don't need to do that any more
  15. Bruce McL

    Bruce McL Mu-43 Veteran

    I have a GM5, I shoot RAW, and I use Lightroom. Importing your RAW file I see the white balance settings are very odd. I have not seen 4400 and +43 in white balance in any of my images. Cycling through the Lightroom presets of Daylight, Shade, etc. I see colors in your image that are reasonably close to what I see doing the same thing with one of mine.
  16. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    Real Name:
    Hey Bruce
    interesting, at work at the moment and ... I just used PS CS6 ACR9 (which I don't have at home) and found the same numbers when it loaded. I also that when I forced it to use "daylight" or "cloudy" it looked the same as I'd expect them to look.
    Last edited: Jan 13, 2016
  17. Bruce McL

    Bruce McL Mu-43 Veteran

    Good to get some confirmation. So it seems like a lot of the camera is working as it should, it's just making an odd decision about WB.
    I would suggest to Zeus1 to check customization settings, then do some resets and firmware updates.
  18. First off, a big thanks to all of You for Your ideas and input.

    @ pellicle: I owe You a big pint whenever ou’re in my neighborhood; learned a lot fromn Your detailed analysis. Will look into dcraw the following days, seems a powerful piece of software. Did not know that dcraw can extract all that EXIF info from the embedded jpg, where Photoshop and DXO Optics Pro find it.

    @BruceMcL: TXS for looking at my image and playing with it. You convinced me that basically there is nothing wrong with my GM5, only some weird software setting.

    I will now check all my settings and try some images with other settings for the white balance, perhaps it’s a problem with the Auto-WB setting.

    Will keep You posted.
  19. pellicle

    pellicle Mu-43 All-Pro

    Feb 10, 2010
    Southport, OzTrailEYa
    Real Name:
    @Zeus1@Zeus1 thanks, glad I could help :)

    a few quick points:
    • draw is command line only, in case that wasn't clear. Its actually the underlying converter for many packages out there that play with raw files. depending how you feel a CLI is either like or hate.
    • draw doesn't extract the EXIF data, it only pulls out the JPG from the raw, To access the EXIF from the JPG, I used irfanview (my preferred image viewing and quick edit tool) .
    • one of the beauties of RAW is that settings such as color balance are nothing more than notes recorded in the RAW file, so you can cook it into anything you want with nothing to lose (unlike JPG) no matter what or settings were :)
    Last edited: Jan 13, 2016
  20. Played some more with test files. Extracted the jpg from the RAW using "InstantJPEGFrom Raw" (I'm on a Mac) (link Instant JPEG from Raw), could read complete EXIF info from it using Apple's preview program. Now, if something basic as Preview can access embedded and extensive EXIF info, why does Photoshop nor DXO do this up front?????