Skip navigation
This discussion is archived
Currently Being Moderated

XZ-1 Orange color not accurate?

Mar 28, 2011 9:36 AM

Hi,

 

I just searched the forum, nobody seems to raise this issue yet.

 

I have an Olympus XZ-1, after I use Lightroom 6.4 RC for a few days, I realize that the orange color developed in RAW is noticibly different from the Jpeg color. At beginning, I thought was a difference due to the saturation and special "olympus color".

 

But after a careful examination, I realize the olympus color is accurate compares to the real thing, the RAW seems a bit off.

 

I took an example to illustrate this.

P3200262 JPEG.jpg

This is the JPEG version, notice the cap of the creme is orange. This color looks very close to the real thing.

 

 

 

P3200262 RAW.jpg

This is developed in Lightroom with default setting, the cap looks yellowish.

 

The difference is quite huge, I need to set the yellow hue to -60 to roughly match the jpeg.

 

 

P3200262 Capture One.jpg

Just for an extra confirmation, the one above is developed from RAW using the default setting of Capture One. The cap color looks the same as the JPEG and the real thing.

 

This happens in all kind of lighting situations, might not be the white balance problem.

Can anyone have a look into this? The orange just looks like yellow.

 

Regards,

 

Luke

 
Replies
  • Currently Being Moderated
    Mar 28, 2011 11:27 AM   in reply to lukebai99

    For Canon and Nikon (and a Leica and a few Pentax) cameras, besides their standard Adobe profile, Adobe makes Camera-matching profiles that attempt to simulate the way the camera JPGs look.  Unfortunately, there are no Camera-matching profiles for Olympus cameras...perhaps a resource-allocation decision based on the perceived number of Olympus users vs the time and expense of creating camera-matching profiles.  Even with the additional profiles the colors aren't always the same, just closer more of the time.

     

    You can see what cameras are supported with the camera-matching profiles by looking in:

    C:\Users\All Users\Adobe\CameraRaw\CameraProfiles\Camera

     

     

    You can try creating your own camera profile for PS and LR using something like an X-Rite Passport:

    http://xritephoto.com/ph_product_overview.aspx?ID=1257

     

     

    What is the WB of the RAW file?  Was there fluorescent lighting in the mix?

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 28, 2011 11:19 PM   in reply to lukebai99

    Luke, try this one

     

    http://www.megaupload.com/?d=C23KRQLU

     

    I made it from Adobe standard, but replaced a lookup table in the profile with version that was used in older Adobe standard profiles (hence the name - Adobe Standard Old). Should have lower orange hue, although not as low as jpeg from camera. Hope it will work, as I don't have XZ-1 to try it. That's all I can do

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 29, 2011 6:17 AM   in reply to Vit Novak

    From my reading, the Adobe Standard conversion in LR isn't orange enough.  Does "lower orange hue" mean more orange or less orange?

     

    Comparing the camera-embedded preview JPG extracted with ExifTool with the Adobe Standard LR conversion of a sample ORF downloaded online, there is clearly a difference in hue for a specific range of colors, and in this example it is the wall behind the clown.  Either Adobe hasn't spend much time with the profile because it is not for a Canon or Nikon DSLR or if their shifting orange to yellow is on purpose, perhaps they are trying to optimize skin colors to something less snookie-like (less orange) but it obviously affects things besides skin:

     

    2011-03-29_075213.jpg

     

    Looking at a RAW+JPG pair file from my own Canon camera, the old-style ACR 4.4 profile is clearly more yellowish and the Adobe Standard is close to the in-camera JPG conversion:

    2011-03-29_001720.jpg

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 29, 2011 11:22 PM   in reply to lukebai99

    My comments about shifting the hue were directed towards Vit who supplied a different Adobe Standard that was modified to use the older calibration matrix, and from my experiment with my Canon Adobe Standard vs the older ACR x.x, the ACR x.x calibration was more yellow and that applied to a too-yellow image would make it worse not better. Did you try Vit's Adobe Standard Old and was it even worse or did it fix things, somewhat?

     

    Whether the ORF or the JPG is more correct depends on the color of the stain of the wood in the background.  I thought the JPEG looked better if it was something like cherry that had been used. 

     

    As far as Adobe caring or not, Eric Chan/MadManChan is the main RAW engineer that chimes in when he feels like it sometimes so if he doesn’t say anything you can imagine that the issue will be ignored but perhaps it'll be addressed silently.  If nothing changes it'll be up to you to create your own profile(s) with something like the X-Rite Color Checker Passport by photographing the standard color target in a couple specific colors of lighting and then using the free program to create your own profile. 

     

    It has been my experience that it is impossible to get a perfect calibration and fixing one color makes another further off, so it's possible that Adobe has done the best they can do with their particular calibration methods and taking another look wouldn't change anything.  Or maybe a mistake was made and things could be better.

     

    To demonstrate the difference in color-rendering a different profile can make, take a look at these experiments I did some years back, using a script Tom Fors created and Tindemans and others modified that would compute the optimum positions of the calibration sliders to minimize the color error in different lighting.  On my slow computer, it would take hours, sometimes, for the scripts to converge on the optimal values.  Adobe releasing their own tool to create profiles made the scripts mostly obsolete and they have stopped being maintained so likely won't work in the current version of ACR, anymore.

     

    http://www.pbase.com/ssprengel/_tech_adobecalibration

     

    The images in this gallery represent two different color-checker photographs that were rendered with various camera profiles and/or optimal slider values, and the distance of between the circle and the square is the amount of error, with the number in parenthesis being the error and the lower is better. The "error" is interpreted as the 3-dimensional "distance" between the RGB value measured from the photograph with the color-profile/slider-values applied compared to the published standard RGB values for each color patch.  For my camera, a custom-created camera profile had the smallest error for the incandescent (2700K) photograph, but the ACR 4.4 profile + Tindemans slider values had the least error for the Hazy Sun (6500K) shot.  The last item on each row is the profile that Adobe tries to make as most like the in-camera JPG as possible.  At the time of my experiments Adobe was just developing their new camera-match profiles so these were the results with the first beta release of them. 

     

    Having the colors be the most accurate isn't the same as having things look the most pleasing in a wide range of photos, so maybe Adobe has perfected their Adobe Standard profile for your camera as much as they want and won't give it any more attention, or maybe someone will see this thread and decide to tweak things.  It is not Adobe's job to make their Adobe Standard profile recreate the in-camera JPG conversion so don't hold your breath. 

     

    My suggestion would be to wait until ACR is released and see if things have improved, and if not, buy a color-checker passport and make your own profile for different lighting situations.  I have my default color-profile as one I've created myself, but sometimes switch to one of the Adobe-created ones if it looks better for a particular situation.

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 29, 2011 11:57 PM   in reply to ssprengel

    I made that profile using Adobe standard for XZ-1, but replaced a lookup table with a table from one of older Adobe standard profiles. So color matrices and forward matrices in this profile are the same as in the profile Adobe Standard for XZ-1. Profiles like ACR 4.4 used only matrices and no lookup table

     

    As I thought, orange hue is slightly lower with this profile (right half of the attached picture), but there are also some differences in red hue, which is also lower.

     

    However, you won't get the same rendering as the camera with this or any other "Adobe standard" profile (in case Adobe changes color matrices in the release version of the profile) - you need a "camera profile" for that

     

    XZ1test.jpg

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 30, 2011 1:46 PM   in reply to lukebai99

    The profiles we can create are based on DNGs from RAW data, only, although you can tweak the sliders after the initial profile computation has occurred to change how things look before saving the profile.

     

    You might ask Vit to see what he did to create the variant of the Adobe Standard profile.

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 30, 2011 11:36 PM   in reply to lukebai99

    Luke,

     

    I'm not familiar with X-rite software or any other widely used calibration software, but I'm affraid you can't get "camera profile" that way.

     

    In my piece of software that I'm using for this I have a color chart with the same number of colors as the lookup table in the profile. And this number is 90x16x16  = 23040, as in latest camera profiles from Adobe, which is way more than those widely used color checker charts. To make a profile, I first have to make a raw file containing this chart with another program, get it developped by Canon Digital Photo Pro or Nikon NX2 and then calculate a profile on the difference between input and output. Unfortunately, I only found a way how to make Canon and Nikon raw files that are recognizable by those programs (using parts of the code found on the internet), so I can't make Olympus camera profiles. I suppose that Adobe is using similar approach, although details of their calibration procedure is unknown

     

    Another approach would be to use camera profile for some other similar camera (with some simple modifications concerning camera model name in the profile). But, it seems that Olympus is using CCD made by Panasonic, unlike Canon G12 and Nikon P7000 that are probably using CCD made by Sony, so its color response is probably significantly different and profile wouldn't work very good

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 30, 2011 11:56 PM   in reply to Vit Novak

    Anyway, try these

     

    http://www.megaupload.com/?d=CY7XU8IN

     

    I combined XZ-1 Adobe standard profile with some camera profiles for Canon G12 (standard and portrait), that are also made by Adobe. On the test photo from dpreview, colors look relatively good to me, so maybe they will be usefull for you in some situations. However, rendering in Canon cameras is a bit different than in Olympus (Olympus generally returning 'warmer' colors than Canon), even under assumption that response of the sensor is the same, which probably isn't. You will also notice differences in brightness and contrast, because of different tone curve than in Adobe standard profiles

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 31, 2011 9:15 AM   in reply to Vit Novak

    Vit, been keeping along with this thread with some interest just out of curiosity. Glad I came across someone as knowledgeable about this as you.

     

    From the work and research you've indicated here, what makes sensors different from one brand to another assuming as you said their response isn't the same.

     

    A CCD is a CCD, so is it the Bayer filter spectral radiance? the electronics such as the A/D converter or is it all made different by the demosaicing algorithm?

     

    Also where do you find this color target with 23040 samples to build the look up table for the "camera profile"? How proprietary is all this camera profiling?

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 31, 2011 7:05 PM   in reply to Tim Lookingbill

    The electronics behind and the filters in front of the sensor can change things even if the photosites, themselves are the same, and there is competition in the sensor market with pressure for  less noise, higher density and lower power consumption/heat-generation, so it would be unlikely that two cameras had the same sensor unless they came out at the same time and only features were different, but Adobe is not a camera manufacturer so it is better they profile every camera.  Some of the profiling has to do with noise-reduction slider settings, and standardization of ISO to brightness, so it's not just colors Adobe has to deal with, and plastic vs metal cases can change the heat/noise characteristics even if the colors are the same.

     
    |
    Mark as:
  • Currently Being Moderated
    Mar 31, 2011 11:38 PM   in reply to Tim Lookingbill

    There is a nice comparison between Canon and Nikon camera regarding spectral characteristics of their sensors and resulting color matrix in the profile

     

    http://www.dxomark.com/index.php/Our-publications/DxOMark-Insights/Can on-500D-T1i-vs.-Nikon-D5000/Color-blindness-sensor-quality

     

    Of course, I don't know the details, but as general rule, sensors with bigger pixels have sharper spectral response of each color channel. I presume it's because color filter in front of the pixel is more effective if the pixel is bigger. Also, I presume there are some differences in materials being used for filter between manufacturers.

     

    Regarding profiles, first I have to point out (although I hope it was evident from my posts) that XZ-1 profiles I attached in this topic are just combined from several profiles made by Adobe, so take them as experimental.

     

    Camera profiles are another story. As I described in some previous topics concerning profiles, I made a program for making camera profiles for Canon compacts that have no raw mode out of the box, but can produce raw files using chdk software. Those cameras are not properly supported by any raw development program, so I tried to make a program to profile them, mostly out of curiosity. One of the functions in chdk software is 'develop raw', so you can take a photo in raw mode and develop this raw in the camera to jpeg later. Or - you can make a raw file with a program, put it into camera and develop it to jpeg. So I came to idea to make a kind of test raw file that contains a test chart and develop it to jpeg with the camera. Comparing input and output, I calculated a profile. It's a simple idea, although implementation took a lot of experimenting. The biggest problem was jpeg compression, which was spoiling the results. So I tried different methods and my final version of test chart looks like this

     

    G9.jpg

     

    It contains millions of colors and covers all points in the lookup table of the profile. This test chart does not exist physically, but just as raw image. Of course, it's not simple like that - various parameters have to be taken into account, like white balance, white level, black level etc ... to produce useful test raw file that can be used to make a profile

     

    Making raw file for those compacts was easy, because it's just binary dump of data, recorded by sensor. Ordinary raw file, produced by cameras that "have raw mode" is a complex tiff structure. But I found some code to make cr2 and nef raw files, so after correcting some misconceptions in that code, I was able to make proper raw files for Canon and Nikon cameras and get them developped by Canon Digital Photo Pro or Nikon NX2, so I profiled my 400D, because camera profiles provided by Adobe several years ago had some flaws. But their calibration method improved significantly in the meantime, as I shown in one topic on this forum, so their "v3" camera profiles are now almost as precise as mine. I hope they'll become even better ...

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 1, 2011 4:36 AM   in reply to lukebai99

    Luke,

     

    all profiles I attached here have, believe or not, the same WB, because I used color matrices from Adobe standard profile (checked once more to be sure). They have different lookup table, and in case of camera profiles, forward matrices and tone curve. However, rice on your photos looks different, because with WB that you used, it's not white in the photo (meaning R=G=B), but a kind of orange color with hue around 40 and saturation around 50. Since we have AWB built in our eyes and brains, it still looks almost white, because other orange colors are more saturated. And, since rice doesn't appear white, it looks different, depending on profile used

     

    You can use WB dropper tool on the rice, that will set WB to rice and it will be about white. In that case, it should look more or less the same with all of these profiles. However, photo will look quite 'cold' in that case, it will loose this orange cast

     

    Of course, it's possible to tune WB of profile by changing its color matrices. It can be done even with DNG profile editor. But, since Adobe team surely has equipments to determine WB properly, I think there's no reason to change it. Also, if you used WB As shot in ACR for these photos, it wouldn't make any difference - photos would look the same with 'wb tuned profiles', only color temperature and tint displayed in ACR would be different. However, if you set WB manually in ACR with those  'wb tuned profiles', WB of resulting photo would be different in that case

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 2, 2011 8:03 AM   in reply to Vit Novak

    Spent some time reading the DXO link and what others here have posted including the past threads Vit participated in concerning his camera profiling methods, a subject that's intrigued me since I edited my first Raw off my Pentax K100D DSLR.

     

    I'm not a programmer, just a former photo realistic painter who's studied the visual effects of light on color. Just can't wrap my head around writing code to manipulate pixels on screen. It's very interesting seeing how someone goes about doing this.

     

    I have a ton of questions mainly concerning tertiary color detail and color constancy manipulation through profiling from a visual rather than mathematical code matrix perspective that's most likely beyond the scope of this thread.

     

    Just want to thank all for their input to my questions and answers to Luke's.

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2011 11:02 AM   in reply to lukebai99

    With the Adobe Standard profile, Adobe's job is not trying to duplicate camera processing in P mode as you say.  It is going for a look that is consistent across all the camera brands and models it has dealt with over the years, and historically probably geared toward commercial magazine and catalog print output where clothing colors being correct for the end consumer are the most important, so using Adobe Standard probably won't produce images like the in-camera JPGs.

     

    For Canon and Nikon cameras and a very few others, Adobe makes some effort to replicate in-camera JPG processing with additional "camera-match" profiles.  Because your camera is from a less common manufacturer, Adobe has chosen not to produce camera-matching profiles for it, and you're left to beg from others or create your own with third-party color-targets and software.

     

    You will find that producing your own color profiles with an x-rite color checker passport may also not match the in-camera JPG, since you're not involving JPGs in the process, only RAWs converted to DNG.  But it should make the colors look more correct as far as the original scene, to the extent that using a 18-color-patches + 6 grays can cover the full spectrum of color, and this correctness may or may not be right for a particular scenario.

     

    I helped with a wedding last summer and dutifully took a few shots of the color checker at various points in the church to get the mixture of daylight, artificial light, and a touch of multi-colored stained-glass coloring things here and there.  However, I ended up using Adobe's Camera-Standard (camera-match) profile for my Canon DSLR, because my own custom profiles brightened the colors too much.

     

    Specifically, the church had neutral-colored walls along the sides, and various colors of artificial high-pressure lighting overhead, and one set of bulbs in particular had an ugly yellowish cast, so wherever this artificial lighting was close to the wall, there was a yellowish area compared to the more bluish diffused daylight elsewhere.  For this particular shoot, it was more important to minimize the saturation of this ugly yellow area than it was the have all the colors correct.

     

    I was able to almost neutralize the yellowish areas by selective HSL adjustments in LR where by using the Camera-Standard profile there was less of this color to remove so it was easier.  Had I not taken so many pictures, and had more time to experiment, I might have found a set of tweaks to the RGB adjustment sliders beneath my own custom profile would have also minimized the yellowish area, but I spent my time perfecting various things for each photo, rather than perfecting the color profile for all of them.  In this photoshoot, having the correct colors was less important than having pleasing photos.

     

    Most of my photos are natural scenes, outdoors so I almost always prefer my own custom profile(s) for these.2011-04-05_125741.jpg

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 5, 2011 12:50 PM   in reply to lukebai99

    I have another non-standard camera brand, Pentax K100D DSLR, that Adobe doesn't include Picture style profiles nor makes an attempt to replicate the jpeg with either Adobe Standard or default ACR 4.4 profiles, but it still does pretty decent job on a wide range of memory colors like orange.

     

    See below the four color variants of orange compared to the Pentax jpeg of a test shot lit under a Philips "Natural Sunshine" 5000K/91 CRI fluorescent tube bought at Home Depot. Exposure specs are...AutoWB, 35mm, 1/30's, f/7.1, ISO 400. All that was changed in the three non-Pentax jpeg was the profile with the forth being a custom DNG dual illuminant profile made from an X-rite ColorChecker chart.AdobeOrangeRendering.jpg

     
    |
    Mark as:
  • Currently Being Moderated
    Jun 25, 2011 11:44 PM   in reply to lukebai99

    Hi Vit, thank you VERY much for providing your modified profile!

    Great improvement at minimal effort!

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 17, 2012 3:21 PM   in reply to Vit Novak

    Would be great if you could upload your profile to somewhere new, as megaupload is now non-functional.  Thanks in advance.

     
    |
    Mark as:
  • Currently Being Moderated
    Apr 18, 2012 12:11 AM   in reply to foolberry

    Here it is, I found it somewhere on the disk ... however, I don't know is it the same sensor as in Canon, made by Sony. If the sensor is made by Panasonic (like in LX-5), I suppose colors will be undersaturated, as that sensor is less sensitive to colors

     

    http://www.mediafire.com/?du2w4c48ypdl8d2

     
    |
    Mark as:

More Like This

  • Retrieving data ...

Bookmarked By (0)

Answers + Points = Status

  • 10 points awarded for Correct Answers
  • 5 points awarded for Helpful Answers
  • 10,000+ points
  • 1,001-10,000 points
  • 501-1,000 points
  • 5-500 points