This content has been marked as final. Show 19 replies
There are a number of reasons, and I am probably not the one to answer, but I'm going to anyway. Many of the software programs provided by the manufacturers take advantage of proprietary data that is hidden from ACR, and they process the images similar to how the JPEG images are processed in the camera. This proprietary data will never be available to ACR.
ACR, on the other hand, takes the raw image data and displays it. Then it provides you with the controls to customize the camera profile so that it displays the colors the way you want them displayed. You might want stronger reds, or deeper blues, etc. So ACR allows you to modify the profile to meet your specific shooting style and your specific needs.
Thank for the answer Jim, I agree with You for the most part but the fact is that making a in-house calibration with a GM panel and using Greg or Fors scripts to calibrate ACR makes always image better compared to the pure RGB values for the color space used.
I would like to have less difference between factory settings of ACR calibration (the profile canned in ACR is camera-related anyway) so that only a small part of users should need to recalibrate to obtain more "pure" colors.
I would like to know why Adobe shots of GM panel (these shots are used to create calibrations for the camera) bring so different results against those that a normal user can obtain from the same shoot he takes.
Probably only Thomas Knoll or his staff can give a final answer.
For now I always give my clients the advice to calibrate ACR using the GM panel. For many I do it as a add-on service. But is more a pain that a gain for me.
Personally, I have found that the results produced by ACR from my Nikon D70, D200, and D3 are not that different from those produced by Nikon Capture (Nikon's raw converter, which gives results very similar to in camera JPEGs) when the camera rendering options are set to neutral.
Nikon Capture reads the camera settings and applies them, whereas ACR ignores settings except for white balance. By neutral, I mean normal contrast, normal saturation, no hue adjustments, and a normal tone curve. The camera factory defaults vary. Cameras for the amateur market tend to have settings favoring saturation and contrast, whereas the professional cameras tend to have a more neutral default. These defaults are usually given in the DPReview writeup for the camera.
Since the OP asked for some ΔEs, here they are. I photographed a MacBeth color checker with the D200 with daylight illumination. Settings were: Color Mode, II; Tone Compensation, normal: Hue Adjustment, 0; Saturation, normal; and Sharpening, none. I then compared the results from ACR and Nikon Capture NX with Imatest, which gives the ΔEs, presented here in graphical format.
Color checker with NC using normal camera settings:
Color checker with ACR default settings:
NC analysis with normal camera settings:
ACR analysis with default settings:
ACR post calibration with Fors Script:
For help in interpreting the charts, the reader should refer to the Imatest web site, which has detailed explanations. In summary, the further from the white point in the center of the plot, the more saturated the color is. Shifts in hue are indicated by shifts in the *a, *b axes. As can be seen from the plots, the ACR rendering is actually slightly more saturated and has overall better color accuracy then the NC rendering. The Fors script calibration gives near "ideal" camera results with ACR.
As Thomas Knoll and others have pointed out, a neutral rendering is often not the most pleasing one, and saturation and contrast are often boosted, but this is a personal preference.
With the Fors script, the ACR settings are set to a very flat image. After calibration, the ACR controls are reset to normal and the calibration settings (in their own tab are left as set by the script).
If you re-run the Imatest analysis after calibraiton with the normal settings restored, the colors are quite saturated and you must talk this into account when using the script calibration.
From all of the reading I have done on this forum, as well as my reading of Real World Camera Raw with Photoshop CS3, it is my understanding that the underlying reason that ACR cannot provide those factories savings is because that is proprietary data that is not available. The proprietary data simply has to be ignored by ACR.
> From all of the reading I have done on this forum, as well as my reading of Real World Camera Raw with Photoshop CS3, it is my understanding that the underlying reason that ACR cannot provide those factories savings is because that is proprietary data that is not available
Jim, it is mainly *you* (and Jeff), who propagated this rubbish. You need to be more selective to which posts you pay so much attention.
The related proprietory data is not available *in pure DNG format*, i.e. without MakerNote, and obviously ACR is acting in a manner to ensure, that the result of a native raw conversion be the same as that from a pure DNG.
In cleartext: ACR CHOOSES not to use that information.
Thanks to Bill for his full test.
However it confirms 2 things (tell me if i misunderstood):
1) ACR settings are always better (more precise) than NX. We all know colors from NX are not so precise and NX itself uses some tone compensation (that cannot be excluded) to "clip" blacks (and somehow hide sensor noise). A linear NX curve is NEVER really linear if compared with the linear settings of ACR.
2) The calibrated (with Fors script in this case but Rags script is for me more balanced) ACR setting have better DeltaE than default ACR profiles (with linear curve and all settings to 0 except brightness and contrast).
The point is why the original profiles have more deltaE than a Fors script (that is also iterative and should be less precise than an optimum profile canned into ACR). Reds are duller regarding NX (for the example) and this is the thing Nikon users note most and hate of ACR. Canon users seems to have similar complaints.
It seems that ACR profiles are quite conservative on some tint and less on others. ACR calibration lack lightness slider (uses only hue and saturation). TK has his good reasons to do this probably assuming lightness is common to all hues.
It is possible to set ACR controls to mimic somehow NX or DPP behaviour of course. But ACR canned profiles seem always "weak" on red hue rendering (and some flash tones) for example wich is a very common complain among NX customers I hear.
ACR have a calibration setting for a good reason. But in my opinion it should be used in rare occasions, not as a "common" procedure.
What surprise me is that the same strong error (reds domain) is present in all camera profiles (I have see at least 20 different camera cases). And this push me to think to a camera profile process error.
I think that hidden data are not a problem in color stuff
You've got a lot of guts to say that what Jeff says is rubbish, but so be it. Consequently, I will bow out of this discussion and leave the answers to those of you who are keenly better informed than I.
> Consequently, I will bow out of this discussion and leave the answers to those of you who are keenly better informed than I
Alternatively, you could look into the matter, though it is certainly more labour intensive than perpetuating something you accept as "wisdom".
I give you an example.
Canon has introduced the option Highlight Tone Priority with the 1DMkIII. The effect of this setting is relating the raw data, that the selected ISO (note: not the exposure) is reduced by one stop (in order to prevent clipping). The effect on the raw processing (in camera and in DPP) is, that the intensity gets increased by one stop, but the very highlights are saved somewhat, comparable to the "recovery" setting of ACR.
When presenting the raw data, I needed to know if HTP has been selected or not. So, I compared the MakerNote of two shots with otherwise equal settings, and found that a particular field (a complex field with sub- and sub-substructures) has become extended by several new values, and one of those is the HTP flag.
You would say now that this is "not available" for ACR, right?
Well, here is the surprize: the support by ACR of the 40D (and I guess of the 1DMkIII and 1DsMkIII) DOES INCLUDE recognition of the HTP setting.
There is a DNG field for automatic intensity correction (exposure correction in ACR's terminology), and that will be set to +1 EV, when the HTP flag is on. However, there is no DNG (or otherwise) field for the "highlight recovery" or "highlight saving" (as nothing will be recovered by this option), so ACR is going the worse way: implementing half of the setting, thereby eliminating the effect of the ISO modification (the worse is, that the user does not see any indication of the automatic intensity correction, but that's a different issue).
Btw, even the correct interpretation of the WB requires using some MakerNote fields.
Finally, a case I find particularly interesting: the Nikon D70 (perhaps other, older models as well) does not record the ISO setting in Exif, where it has a proper representation; it is in the "secret" MakerNote. And you know what, Jim? The ISO number DOES APPEAR in ACR.
There, that must have made you feel much better. I didn't know that I didn't know that the ISO setting DOES APPEAR in ACR. But then, since I don't have a Nikon camera, I'm not in a position of knowing anything, right? Rant on. It makes no difference to me.
you are being either obtuse or your English comprehension is really low. The point of the example with the D70 is, that THE ISO SETTING HAS TO BE TAKEN FROM THE "SECRET" INFORMATION, which is, according to you, not available to ACR.
There is no need for shouting or blaming, Jim is just referring to the common use of ACR. Thomas Knoll provided a very good starting point for each cameramodel. You can make your own standard setting in a splitsecond for each model or situation.
There may be some need to write over and over again that some color in some situation is off, so be it.
Many of this forum users and readers just take this starting point to create whatever they want. And that is the way it should be. There is no standard situation, every image can be made as the creator wishes and that is what makes ACR such a wonderful application
I find it very disturbing, that the average level of reading comprehension of the posters on the Adobe forums is comparable to that of elementary school children.
Does not it disturb you, Omke, the very least, that I did not post a single word here relating to the quality of the ACR conversion? I personally never complained about ACR not using the in-camera settings, for all my settings are always neutral in order to achieve a raw-like in-camera histogram.
For your understanding: the subject I posted to is, *why not* ACR is using the in-camera settings. Furthermore, I did *not* post on this thread, *why* ACR is not doing so in my opinion; I only disspelled a myth, which has been spread repeatedly on this forum.
Hi to all
I think the theme has been interpreted by some forum users in the wrong way. Or I haven't been clear enough. I make my apologies for this.
The original keypoint was: why the camera profiles created by Adobe are so different from a calibration profile that the user can create by himself using the same GM 24 patch panel and a script? What is the process used by Adobe to create the calibration for a camera? How can be this process so different (and give so different colors) from a manual calibration made afterwards by the user?
No problem with some optimizations and curves made by brand softwares, that's normal. The point is why the reds and blues (to pick the more visible ones) of canned Adobe profiles are so "dull". Are GM 24 patch panels used by Adobe different from those used by normal ACR users?
Even if ACR reads some of Makernotes for some cameras, this is not correlated to the pure color render of the profiles made in ACR.
>The original keypoint was: why the camera profiles created by Adobe are so different from a calibration profile that the user can create by himself using the same GM 24 patch panel and a script? What is the process used by Adobe to create the calibration for a camera? How can be this process so different (and give so different colors) from a manual calibration made afterwards by the user?
This topic has been discussed at length on the Luminous Landscape forum:
Thank You Bill, I read the full discussion on LL. The initial reasons said by Schewe are however well know (more samples from more cameras bring different results).
From the full discussion I haven't however found the true reason why the ACR factory calibration setting should be considered 1) a mix of more GM 24 patches to give a pleasant overall rendering or 2) a mix of first-out cameras from vendors who send uncalibrated or very different specimen to Adobe or even 3) a mix to compensate for strange "in-house" settings that camera makers do on their models.
The fact against point 1) is that for ALL cameras are weak on red. Ho can be this color the only one that "damage" all the others so it must be "dull" to give good skintones or grass?
The fact against point 2) is that proprietary makers sell only ONE software without special setting and it performs the same for all samples (I don't think it reads camera serials....). Even more the profile can be updated with more samples in the camera lifetime so the resulting profile should cover the "mean" of most.
The fact against point 3) is that even with a "generic" software like ACR it should be quite easy to mimic other sw behaviour.
The fact remains that without re-calibration the most part of ACR profiles for mainstream cameras are considered "weak" and discarded.
In a open market world, could Adobe create a NX-like profile? Or a DPP like profile for a camera? Only for color render of course!
The mistery remains for me. Ok let's go calibrate for a new client camera....
>The fact remains that without re-calibration the most part of ACR profiles for mainstream cameras are considered "weak" and discarded.
While you raise an interesting point, at least according to my tests, ACR has a smaller ΔE than Nikon Capture NX when ACR is used with default settings and NX with "normal" camera settings. I find the default ACR profile satisfactory for most of my work and do not agree with your assessment.
ΔE is the Euclidean distance between L*a*b* values of two colors in the 3 dimensional space. This error contains chroma (saturation), hue, and luminance components. Saturation is often boosted for a more desirable appearance and the ΔE from saturation is not necessarily undesirable, while shifts in hue are generally not desired. On the Imatest graphs, ΔC represents chroma error and values are given corrected and uncorrected for this saturation boost. See the Imatest documentation for details:
Jim Hess wrote:
>> it is my understanding that the underlying reason that ACR cannot provide those factories savings is because that is proprietary data that is not available. The proprietary data simply has to be ignored by ACR. <<
Agreed in parts.
... ever thought about the time zone setting of your camera:
I think the question is less about any secret information hided by camera manufacturer, but more about the frame parameters which ACR wants to take into consideration to tweak its internal math and transforms.
Lets take for example the Vibrance slider. As far as I understand, it provides a quite advanced definition for Saturation which effects less saturated colors more than already high saturated ones. Also it seems to mask skin tones via the hue angle. But it doesnt go so far to make an area-based selection i.e. face detection of the pixels to build such mask. Or, it wont tweak the hue of skin tones depending on the time zone and language settings of your computer (i.e. cultural background of the user).
That said, such high end details might have less to do with the issue described by the thread opener. There seems to be some confusion about the Accuracy of profiling, the merits and pitfalls of the Fors script, and what is called a Preferred rendition. Different chapters, and I recommend to follow Bill Janes posts on this.
My 2ct - just trying to comment beyond elementary school level:)
>ACR has a smaller ΔE than Nikon Capture NX when ACR is used with default settings and NX with "normal" camera settings. I find the default ACR profile satisfactory for most of my work
Yes Bill You are right. NX has imatest result far less correct than ACR. I replicated them with ease. The ACR canned profiles are probably calibrated by TK in a form that privileges the skin tone (in some ways) and makes red appear "duller". Normal calibration with Rags script (that makes a weighted evaluation on more patches and gives also the option to use a different patch for calibatichg reds) is much more "reddish" and gives to skin a red tone. Calibrating using other patches for reds (instead of the red one) gives results similar to ACR defaults. Working manually is however possible to mediate to obtain a better calibration.
Probably the intervention range of the three calibration sliders is unable to calibrate correctly for some critical patches on the CC 24 panel.
I agree that saturation and delta can be misleading and that for now is diffcult to find unique calibration settings that can be used for every kind of image.
The fact remains that without tweaking on some other ACR control other than calibration is difficult for the normal user to have a nice red on most of his images. May be that ACR calibration should evolve on something more that RGB hue and saturation.