Pages: [1] 2   Go Down

Author Topic: Wrong profiling for the Sony RX100?  (Read 9783 times)

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Wrong profiling for the Sony RX100?
« on: November 05, 2014, 07:02:26 am »

I'm in the process of evaluating C1 v.8 for my Sony RX100 camera and I've run into a couple of problems.

1. The RX100 is a 20MPix camera -- the file dimensions in C1 output recipe summary are the same as outputs from LR5 and DxO Pro 9 (with its automatic lens distortion both off and on): 5472x3648 px. I assume that these are the same dimensions as the camera-created jpeg. However, when you output those files as jpegs or tiffs with C1 the dimensions seem to be uprezzed to 5731x3821 and also the output recipe summary now shows these uprezzed dimensions for the raw file. What's going on here? There is no such discrepancy for my Pentax K10D so I assume it's got something to do with some internal output recipe for the RX100 (?). Wouldn't such an uprezzing be detrimental to output sharpness of the file?

2. The lens corrections profile for the original RX100 has the Light Falloff slider permanently off and doesn't correct for the vignetting of the lens. You can access the slider only when you manually switch the profile to the RX100 Mk II lens (it's seemingly the same optics hardware-wise). Also it seems to undercorrect or overcorrect the distoritions at various focal lengths. That second profile makes the corrections similar to what you get from the in-camera jpeg, LR5 and DOP9, so this must be a profiling bug for RX100 Mk I.

3. The colours are wrong. When I shoot the ColorChecker Passport the red patch is orangish (there's too much yellow injected to red/pink tones, possibly to make skin tones more pleasing -- this might work with Canon, Nikon or Pentax reds but it doesn't work for Sony, at least not for my camera). So, the Caucasian skin rendition is sub-optimal and generally greens and yellows get too dense. You have to resort to the Color Editor to neutralize things and start optimizing the file only after you've corrected the defaults. Also, the shadows have magenta cast to them. The easiest way to correct for this second problem is changing the input shadow values of Red and Blue channels by 1 and 2 units respectively in the Levels tab -- you can make a preset for that -- but you shouldn't have to do that for a well-profiled camera.

4. A good point about C1: the LCC tool. The RX100 lens is prone to lens colour casting (esp. at the widest and at the longest focal lengths). Some of the magenta cast mentioned in point 3 can be exacerbated by this phenomenon. C1 is the only converter I know of that allows a semi-automatic way of correcting for that so kudos for that.

So, are there any other Sony RX100 C1 users in this forum and have you encountered similar problems? I've raised a support ticket with the colour issue but it didn't go anywhere and the case was closed with the conclusion that everything is within the norm. All in all, C1 makes me hate my RX100, which cannot be said of LR5 or PhotoNinja, which do very well with this camera. As it is I can't see myself ever reaching for this converter for this camera...
Logged

Manoli

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2305
Re: Wrong profiling for the Sony RX100?
« Reply #1 on: November 05, 2014, 08:27:15 am »

I've just had a very quick look at the points you raise.

It's correct that the lens corrections profile for the original RX100 has the Light Falloff slider permanently off . I don't know if that's a bug, as you put it, or simply that we've moved on and are now on MkIII.

The lens correction tab has 3 profiles for the camera. Generic, Generic pincushion distortion and the C1 RX100 profile. For the Mk II, there's a 4th - Manufacturer profile. If you look at the thumbnail in the filmstrip, you'll see that the amount the image is cropped varies according to the profile selected. (presumably using the lens correction data embedded in the raw file).

Irrespective of which profile is selected, C1 shows 5472x3648 px and on my setup outputs the same size (only checked a tiff). Suggest you check your output recipe settings and also the 'movement' tab in the LC tool - make sure it's zero'd.

Can't comment on colour, I haven't used the ColorChecker Passport with it, and I'm not the person to speak to about colour in this forum !


Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: Wrong profiling for the Sony RX100?
« Reply #2 on: November 05, 2014, 10:29:19 am »

Thanks for the verification. It made me triple check my settings and I've discovered some mistakes I made before.

As to my point nr 1 this was my mistake -- I thought I'd reset everything but for some reason the crop tool didn't reset its settings. I used the crop tool (together with unchecked "Hide Distorted Areas" in the Lens Correction tool) because it appears there's more to be gained from the file if you re-crop it after applying the distortion corrections. Thus at wide angle it appears the camera can yield a larger FOV than the advertised 28mm (especially if you use a 16x10 aspect ratio). I don't know why the output recipe summary showed original dimensions where in fact the crop tool was instructed to enlarge. So, C1 doesn't "uprez" the file as I claimed in my OP.

My point nr 2 stays. I do think it is a bug. You shouldn't have to change the defaults in this case and actually it takes a bit of experimentation to find out that there is a problem. It so happens that MkII probably has the same optics as MkI, but MkIII is a completely different beast.  There's a workaround for it -- if you've got the original RX100 make a preset which switches the lens correction profile to Mk II and turn on both Sharpness and Light Falloff to 100%. But if you want to use the LCC tool it's probably better not to do this -- the tools seem to be in some kind of conflict if you change the Lens Correction defaults. BTW, I can see only the same three options under the "Recommended lenses" menu if I change the profile to MkII.

I hope somebody can address the colours issue. I discussed this in the Sony DPR forum and two people confirmed this, however it seems quite strange that there so few of us who find the colour rendition problematic. My Pentax K10D renders beautifully in C1 and generally colours is one of the strongest fortes of C1, but sadly not in the case of RX100, at least not for me. Maybe I'm magenta-biased (being a Pentaxian I think I might be).
« Last Edit: November 05, 2014, 10:35:21 am by sankos »
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: Wrong profiling for the Sony RX100?
« Reply #3 on: November 05, 2014, 11:13:02 am »

3. The colours are wrong.

it is a matter of your personal taste (because you did not show any measurements, only what you say you see and your subjective /again - no measurements, that is with dEs in hand/ opinion might be different from others), so use C1 color editor to adjust colors and save corrections as your new camera ICC profile and use it by default or rally troops to show a noticable amount of people not happy

Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: Wrong profiling for the Sony RX100?
« Reply #4 on: November 05, 2014, 11:14:46 am »

however it seems quite strange that there so few of us who find the colour rendition problematic
well, may be this is indeed the case ? there are a lot of variables - so unless P1 is shown numbers, you need to show in numbers
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: Wrong profiling for the Sony RX100?
« Reply #5 on: November 05, 2014, 11:35:38 am »

Quote
it is a matter of your personal taste (because you did not show any measurements, only what you say you see and your subjective /again - no measurements, that is with dEs in hand/ opinion might be different from others)

Right, I get it that it's subjective, but it becomes more objective when I set a well-established point of reference. In my case the point of reference is the physical object (ColorChecker Passport), my calibrated monitor and the same pair of eyes and mind that compare the object with its reproduction (I studies aesthetics so I know something about the subjectivity of colour). I also use other converters to compare and Capture One in this case is clearly an outlier (DOP 9 does kind of OK, LR5 gets a custom-made dcp and it satisfies my need for colour accuracy, RT 4.2.1 the same as LR, PhotoNinja is good out of the box with the RX100).
[edited to delete wrong colour read-outs: will revisit that]
« Last Edit: November 05, 2014, 11:41:17 am by sankos »
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: Wrong profiling for the Sony RX100?
« Reply #6 on: November 05, 2014, 04:32:42 pm »

my calibrated monitor and the same pair of eyes and mind that compare the object with its reproduction (I studies aesthetics so I know something about the subjectivity of colour).

that's not an argument... P1 does not see your monitor and how well calibrated it is (along with the viewing environment where it is located) and does not have your eyes... you want go this route ? measure your target (spectrophotometer), get a shot in raw, present the differences between C1 output and what you think the output shall be in a numeric way (dEs) for patches that you are not happy with... so that your monitor and your eyes are not the part of equation... if you want to argue with "eyes" in hand - then P1 shall see a lot of people unhappy w/ the color transform... once they see a sufficient amount of complaining people somebody might actually take a look.



Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: Wrong profiling for the Sony RX100?
« Reply #7 on: November 06, 2014, 04:44:02 pm »

measure your target (spectrophotometer), get a shot in raw, present the differences between C1 output and what you think the output shall be in a numeric way (dEs) for patches that you are not happy with...
That was not the intention of the original post. At this point I don't feel like being a beta tester for P1, buying an expensive tool and investing some serious time to mathematically prove a point (the more so that I'm evaluating the software before making the decision whether it's worth it to upgrade from the Express version I've been using since C1-4). For me photography is an artistic hobby and I don't require mathematical precision when it comes to colour. Indeed, I tended to gravitate towards C1 because I liked the tweaked colour rendition for my previous camera, fully aware that it was not neutral. My question was directed at RX100 users about their subjective impressions connected with the profiling of that camera.

Quote
once they see a sufficient amount of complaining people somebody might actually take a look.

Yes, they are only bothered if people make some noise: that's the secret formula that gets repeated on the Phase One fora and when you raise a ticket for an unsupported camera (I did that a couple of times and was always answered this way). Do they really want their customers to beg them for improvements of their software?? I always found it lame and I don't think this kind of attitude is a good business model. But what do I know...
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
ColorChecker red patch rendering
« Reply #8 on: November 06, 2014, 05:05:42 pm »

I've compiled 7 renditions of the red patch from the X-rite ColorChecker Passport. All were rendered from the same .arw raw file at the default settings for a given converter, the WB setting was As Shot (I manually set the WB in camera at the time of capture). All renditions were exported as 8-bit Adobe RGB tiffs and combined into the attached file [edit: sorry had to change it to a jpg because the tiff wasn't accepted]. Can you see the odd one out?

For most squares the Green component is around 20 points lower than the Blue. However, squares 1 and 2 diverge from that formula and green gets more favoured. I hope that you can see the difference even without referring to the numbers?
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
legend
« Reply #9 on: November 06, 2014, 05:23:42 pm »

Square no. 1: C1 v.8 at its defaults for the RX100 (Mk I). Lowest brightness. Green much above Blue.
No. 2: DxO Pro 9, the camera default colour profile. Contrasty tone curve. Green and Blue about equal.
No. 3: DxO Pro 9, neutral colour/neutral tone ver. 2 colour profile.
No. 4: LR5, Adobe Standard colour profile.
No. 5: LR5, X-rite generated custom profile.
No. 6: RawTherapee 4.2.1, camera profile.
No. 7: RT, X-rite profile.

edit to add: in CMYK terms there's too little magenta and too much yellow when compared with Adobe, Xrite, Dcraw and DxO neutral renderings.
« Last Edit: November 07, 2014, 04:29:09 am by sankos »
Logged

Rhossydd

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3376
    • http://www.paulholman.com
Re: Wrong profiling for the Sony RX100?
« Reply #10 on: November 07, 2014, 09:34:41 am »

At this point I don't feel like being a beta tester for P1, buying an expensive tool and investing some serious time to mathematically prove a point (the more so that I'm evaluating the software before making the decision whether it's worth it to upgrade from the Express version I've been using since C1-4). For me photography is an artistic hobby and I don't require mathematical precision when it comes to colour. Indeed, I tended to gravitate towards C1 because I liked the tweaked colour rendition for my previous camera, fully aware that it was not neutral. My question was directed at RX100 users about their subjective impressions connected with the profiling of that camera.
As a RX100 and CO8 owner I'd agree that the default colour rendition is pretty poor.

Subjectively; over saturated and too red.
Mathematically it's more complex, but still the overall results confirm the subjective opinions here.
Average/Median value/Standard deviation
Lightoom default----29/25/14
Capture One 8---38/36/17
Lightroom with QPcard profile---27/24/11
DXO 8---23/22/8

The practicality is that you can improve these numbers by careful set up, with custom saved setting in any of the programs. However the defaults are clearly better on DXO 8 with Lightroom using a custom camera profile second and Capture One 8 being the worst.

For me, I bought the upgrade to Capture One 8 because it was so cheap. I like supporting some competition for Adobe and CO does have a few useful features for working colour selectively that the other programs don't have. It won't be my default RAW converter as overall Lightroom does 99% of what I need and overall gives better results than the others I've tried, plus it's got by far the best workflow and usability of any of the competition.

Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: Wrong profiling for the Sony RX100?
« Reply #11 on: November 07, 2014, 10:12:29 am »

Yes, they are only bothered if people make some noise

that is the case with everybody - typically if you want something you need either to raise a sufficient amount of noise or really prove the case with numbers in hand, but certainly there are exceptions to every rule

But what do I know...

well, when P1 decided to provide a special (cheaper, Sony locked) Pro version for Sony owners, then also added some extra profiles in C1 for some recent Sony cameras (like A7, 7r, 7s) but not for RX100s... so may be they will expand the list of Sony cameras for which their own alternative profiles will be available
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: ColorChecker red patch rendering
« Reply #12 on: November 07, 2014, 10:16:42 am »

All were rendered from the same .arw raw file

may be you can post that raw file of yours and some people will chip in ?
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: Wrong profiling for the Sony RX100?
« Reply #13 on: November 07, 2014, 02:06:46 pm »

As a RX100 and CO8 owner I'd agree that the default colour rendition is pretty poor.

Thanks for the sanity check. I thought it was only my camera that was giving me such grief. I still can't believe how much magenta lens colour cast there is in those wide angle photos. Also the 100mm is not at the Zeiss' best...


For me, I bought the upgrade to Capture One 8 because it was so cheap. I like supporting some competition for Adobe and CO does have a few useful features for working colour selectively that the other programs don't have. It won't be my default RAW converter as overall Lightroom does 99% of what I need and overall gives better results than the others I've tried, plus it's got by far the best workflow and usability of any of the competition.

I like C1 although it's missing some of LR's functionalities that I use (e.g. geotagging). It's this colours business that stops me from upgrading because if I don't sort it out to my liking I don't want to waste my time on endless comparisons between various raw converters. I'll try tweaking the Colour Editor and doing a custom ICC. Do you have any hints as to the settings that work for you?
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: Wrong profiling for the Sony RX100?
« Reply #14 on: November 07, 2014, 02:10:11 pm »

so may be they will expand the list of Sony cameras for which their own alternative profiles will be available
Fingers crossed  ;) I'm also curious how much further this cooperation will go. When I heard about it I thought "Kudos to Sony for not choosing Silkypix". I didn't see it coming.
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: ColorChecker red patch rendering
« Reply #15 on: November 07, 2014, 02:12:35 pm »

may be you can post that raw file of yours and some people will chip in ?
Right, this link should work, I think.
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: Wrong profiling for the Sony RX100?
« Reply #16 on: November 07, 2014, 03:41:30 pm »

how about this crude profile, attached
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: Wrong profiling for the Sony RX100?
« Reply #17 on: November 07, 2014, 05:03:07 pm »

Now, that's more like it. If C1 shipped with a profile like that I wouldn't complain, it's a good starting point, thanks! Was it just Color Editor adjustments or did you do any other tricks?
Logged

deejjjaaaa

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1170
Re: Wrong profiling for the Sony RX100?
« Reply #18 on: November 07, 2014, 05:10:57 pm »

Now, that's more like it. If C1 shipped with a profile like that I wouldn't complain, it's a good starting point, thanks! Was it just Color Editor adjustments or did you do any other tricks?
well that was crudely done from your raw, using rawdigger & makeinputicc argyll gui frontend...  gamma 1.8 & matrix (it is not Lab cLUT profile - so coloreditor will not work in C1)
« Last Edit: November 07, 2014, 05:13:34 pm by deejjjaaaa »
Logged

sankos

  • Newbie
  • *
  • Offline Offline
  • Posts: 36
  • Why matters
Re: Wrong profiling for the Sony RX100?
« Reply #19 on: November 07, 2014, 05:15:15 pm »

Oh, OK, that goes beyond my skills, so thanks once again. It looks like there is some hope for me using C1 after all.
Logged
Pages: [1] 2   Go Up