ARGB vs PPRGB in Workflow

<p>Hello All<br>

My workflow is : Convert RAW to TIFF in DXo then use LR3 to fine tune / crop / print. Print output is generally to an Epson 3880. All of this is being done in WinXP pro.<br />At present I have DxO set to export its TIFF images in a ProPhoto RGB colour space using ProPhoto.icm (Copyright Eastman Kodak 2003) which I downloaded from Dodgecolor. LR3 is set to use a ProPhoto working colour space.<br />I seek advice with respect to maintaining colour spaces in these various environments, specifically:</p>

<ol>

<li>Is the RAW (.CR2) image coming out of my Canon 5DMkII encoded with a colour space? I have the camera set to ARGB but I wonder if this applies to raw images or just JPEGs?</li>

<li>I understand that as a general rule of thumb you can go from a large colour space (PPARGB) to smaller one (ARGB) but not the other way around; Is this correct? </li>

<li>So if the RAW image file indeed comes out of the camera in ARGB raw, why not set the whole workflow in ARGB instead of ProPhotoRGB?</li>

</ol>

<p>All comments welcome. Thanks in advance</p>

<p>Raw files do NOT have a color space.</p>

<p>1_just jpeg.</p>

<p>2_yes</p>

<p>3_the raw dont come out in anything.. it become what you set in your raw development software.. it is by default, create with the largest color space your camera can produce... let say ; )</p>

<p>Read this thread on the advantages of <strong>editing</strong> Raw images in 16bit ProPhotoRGB:</p>

<p>http://www.photo.net/digital-darkroom-forum/00VegT </p>

<p>Raw files kind of have a color space (all raw converters assume something differently unless the spectral sensitivities of the chip are defined). Then the converters have their own processing color space which isn’t the same. In the case of ACR and Lightroom, that processing color space is ProPhoto RGB using a linear encoding. ProPhoto uses 1.8, LR/ACR 1.0. So even if you hand LR a ProPhoto doc, its going to process it using its ‘<em>hybrid</em>’ color space. Keep in mind that when you hand LR or ACR a TIFF from Dx0, its no longer raw data (even if you save it out as a DNG)!</p>

<p>The raw file from your camera doesn’t come out as Adobe RGB or ProPhoto RGB. But each converter assumes different things and processes the data differently. The key to ProPhoto is its the biggest container you have to work with from the raw data and converter. You may want to read:<br />http://www.adobe.com/digitalimag/pdfs/phscs2ip_colspace.pdf</p>

<p>All above are valid answers. There is an additional issue anyway: When you output Tiff (or Jpeg) from DxO Optics Pro, the color space used by the program is AdobeRGB. If you select ProPhotoRGB as your output color space, the color will be encoded to that color space but there will not be any color outside of the AdobeRGB color gamut.</p>

<p>If you are using a DxO + LR workflow, the other option is to output DNG files out from DxO and work from there in LR. As Andrew Rodney pointed, they are no longer RAW files. They are linear where demosaicing and interpolation has been applied. They however, could be adjusted for white balance and camera profile in LR as if they were RAW.</p>

<p>Some users complain about muddy colors going this route. In my experience it is because they don't reassign a camera profile once in LR.</p>

<blockquote>

<p>When you output Tiff (or Jpeg) from DxO Optics Pro, the color space used by the program is AdobeRGB. </p>

</blockquote>

<p>It kind of sucks you can’t over-ride this (ProPhoto would be preferable to many). </p>

<blockquote>

<p>If you select ProPhotoRGB as your output color space, the color will be encoded to that color space but there will not be any color outside of the AdobeRGB color gamut.</p>

</blockquote>

<p>What? So its an Adobe RGB processing and just clips everything outside that gamut? </p>

<blockquote>

<p>Some users complain about muddy colors going this route. In my experience it is because they don't reassign a camera profile once in LR.</p>

</blockquote>

<p>A camera (DNG) profile or Adobe RGB? </p>

<blockquote>

<p>What? So its an Adobe RGB processing and just clips everything outside that gamut?</p>

 

</blockquote>

<p>Yes, this can be verified by doing this: Use an image with strong colors and deliberately increase saturation to maximum so that you have all channels clipped. Output a Tiff file 16 bits using ProPhotoRGB. Plot (3D) the gamut of the resulting file against AdobeRGB and against ProphotoRGB. You'll see that the gamut of the file, while encoded in ProphotoRGB, is inside AdobeRGB</p>

 

<blockquote>

<p>A camera (DNG) profile or Adobe RGB?</p>

</blockquote>

<p>A camera profile (DNG)</p>

<p>Thanks to all above; a few comments for further discussion:<br>

Francisco:<br>

Does your comment about DXO outputting in ARGB hold even when the ProPhoto ICC is set in the output template?<br>

Anyone:<br>

Is it correct to say that if all the work is done in LR (meaning the raw file is opened directly in LR), then LR derives its output from that raw file. For example if you crop, white balance and S-curve a raw file in LR and then export it to a JPEG from LR, is that JPEG generated by applying your edits directly to the raw file in the conversion process?</p>

 

<p>David,<br /> Yes, the comment holds when ProPhoto ICC is set in the output template. I used to work that way until I encountered this comment from a thread in the <a href="http://www.luminous-landscape.com/forum/index.php?topic=48153.msg418213#msg418213">Luminous-Landscape</a> forum: (this is claimed to be an answer from DxO support to a customer)</p>

<blockquote>

<p>"DxO Optics Pro converts to AdobeRGB as the output color space. So converting to ProPhotoRGB for Output does not provide any gain. ProPhotoRGB RGB is about 50 percent bigger than AdobeRGB. ProPhotoRGB contains colors outside the AdobeRGB color space, namely in the orange and yellows. ProPhotoRGB covers all of these colors, and also some that are not even visible, let alone printable.<br />Making 16-bit ProPhotoRGB TIFFs can, however give you issues later. The color space is so big that even in 16 bit mode, you risk banding and artifacts due to the "gaps" between the colors. If you were making 32-bit TIFFs and DxO converted to ProPhotoRGB as an output color space, and also editing in 32-bits in Photoshop this might be feasible, but we do not recommend it for 16-bit color operations.<br /><br />Presently you cannot do what you are asking. Thre are no workarounds. DxO's working color space is preset and you do not have access to change it. If you were to make output files in ProPhotoRGB, 16-bit being the maximum bit depth available to you. You would very likely have the problems cited above using them afterwards. You are fooling yourself by attempting this and believing you have gained something."</p>

</blockquote>

<p>After that I did some tests (<a href="http://www.luminous-landscape.com/forum/index.php?topic=54284.0">here</a>) using DNG as output from DxO. This way you get all the colors, you can even do your own DNG profile if using a ColorChecker.</p>

<blockquote>

<p>Is it correct to say that if all the work is done in LR (meaning the raw file is opened directly in LR), then LR derives its output from that raw file. For example if you crop, white balance and S-curve a raw file in LR and then export it to a JPEG from LR, is that JPEG generated by applying your edits directly to the raw file in the conversion process?</p>

 

</blockquote>

<p>Yes, that's the way LR (and ACR) works</p>

<p>Just keep in mind that whatever preview in an image you generate whether default or edited by user in any Raw converter whether you saturate colors to clipping point or not has its color gamut (of that image's scene) limited to the gamut of the display. Have an sRGB-ish display? That's your limit. AdobeRGB-ish display? You have more color options for pushing the saturation/hue.</p>

<p>This is why I can't understand how you can derive an accurate 3D color gamut model shape measuring an image's color in tiff form if that's whats being done as I understand Francisco's entry...</p>

 

<blockquote>

<p>Plot (3D) the gamut of the resulting file (tiff of image) against AdobeRGB and against ProphotoRGB.</p>

</blockquote>

<p>Also what gets confusing in all this is KNOWING for certain the actual source color space the Raw converter is using to generate the default preview which is suppose to utilize the display profile in a traditional ICC connection space transform as Photoshop and other color managed apps implement.</p>

<p>As Andrew pointed out ACR/LR's source is a linear version of ProPhotoRGB whose spectral colorant properties offer a big enough fit for almost any type of scene thrown at it whether or not ACR/LR knows anything about the spectral properties of the camera's RGGB filtering which is the true source defining the color the camera captured.</p>

<p>What now limits the Raw preview and all possible colors available in defining it is the color gamut of the display.</p>

<p>Just for clarity here :<br>

I beleive the both my monitor (NEC PA241) and my printer (Epson 3880) have sufficent spectral response to cover the entire ARGB space. If anyone disagrees I'd be pleased to learn more about that assumption.</p>

<p>After you calibrate your PA241, you can plot its gamut over Adobe RGB (1998) with the SpectraView II software. They should be pretty close. The 3880 has a larger color gamut in some areas than Adobe RGB (1998) but due to the large differences in shape between working spaces and output color spaces, you really need to map the two 3D to see the full picture. </p>

<blockquote>

<p>I beleive the both my monitor (NEC PA241) and my printer (Epson 3880) have sufficent spectral response to cover the entire ARGB space. </p>

 

</blockquote>

<p>While the display can contain a large percent of Adobe RGB (the specs say 98.1%), the Epson 3880 can NOT print ALL of the colors of Adobe RGB but can, ironically, print some color that can't be contained in Adobe RGB. The 3880 can print some oranges, reds and yellows that would be contained in PhotPhoto RGB but not in Adobe RGB. As Andrew says, the gamut shapes are a bit different and you can tell by looking at the profiles in a profile utility in 3D.</p>

<blockquote>

<p>Just keep in mind that whatever preview in an image you generate whether default or edited by user in any Raw converter whether you saturate colors to clipping point or not has its color gamut (of that image's scene) limited to the gamut of the display. Have an sRGB-ish display? That's your limit. AdobeRGB-ish display? You have more color options for pushing the saturation/hue</p>

</blockquote>

<p>The display does not matter here. You can even do this excersise with a monochrome display. A gamut plot of a color space is a mathematical model. Otherwise it would be impossible to plot the gamut of ProPhotoRGB, since no monitor could display it (it will never happen).</p>

<p>There are a few tools to do this. In this specific case I used the tools included in <a href="http://www.argyllcms.com/">Argyll CMS</a>, namely:</p>

<p>iccgamut: to create a gamut file of a ICC profile<br>

tiffgamut: to create a gamut file of a TIFF image<br>

viewgam: to convert one or more gamut files to a 3D visualization file</p>

<p>These 3D files could be then displayed interactively with a tool like <a href="http://www.cortona3d.com/Products/Viewer/Cortona-3D-Viewer.aspx">Cortona 3D viewer</a></p>

<p>What subject or colored objects were in the tiff image you used and saturated to clipping that you could derive a 3D gamut plot from. Did this tiff image contain all the colors reproducible by ProPhotoRGB or AdobeRGB?</p>

<p>When you go full saturation using the tools in ACR/LR you only have 255 levels of 8 bit color preview. At some point during the clipping/full saturation you've plumbed the 255 head room and now into undefinable colors.</p>

<p>IOW how did over saturating to the point of clipping give you something to measure in a tiff image to get an accurate 3D gamut plot? What was in the image?</p>

"Is the RAW (.CR2) image coming out of my Canon 5DMkII encoded with a colour space? I have the camera set to

ARGB but I wonder if this applies to raw images or just JPEGs?"

 

Encoded in the sense that the metadata for the raw file will tell Canon's raw processor that you had the camera set to

Adobe RGB(1998) and to use that as a suggested working space but as raw files do not have color spaces or profiles

then it is only a suggestion. However your camera will use the color space you choose for in camera JPEG production

for the camera's histogram and preview.

 

"I understand that as a general rule of thumb you can go from a large colour space (PPARGB) to smaller one (ARGB)

but not the other way around; Is this correct?"

 

Well you can but there's no real point to it.

 

"So if the RAW image file indeed comes out of the camera in ARGB raw, why not set the whole workflow in ARGB

instead of ProPhotoRGB?"

 

The working space / color choice is only "burned into" the photo when you tell your raw processor of choice to create

a TIFF, PSD, or JPEG. The raw file remains "raw".

All comments welcome. Thanks in advance

<blockquote>

<p>What subject or colored objects were in the tiff image you used and saturated to clipping that you could derive a 3D gamut plot from. Did this tiff image contain all the colors reproducible by ProPhotoRGB or AdobeRGB?</p>

</blockquote>

<p>Flowers with saturated colors (reds) could work, even a colorchecker if you increase saturation to maximum. The image does not necessarily contains all the colors reproducible by ProphotoRGB or AdobeRGB. Just the fact that some colors are outside AdobeRGB will suffice.</p>

 

<blockquote>

<p>When you go full saturation using the tools in ACR/LR you only have 255 levels of 8 bit color preview. At some point during the clipping/full saturation you've plumbed the 255 head room and now into undefinable colors.</p>

</blockquote>

<p>When you reach 255 you just hit the walls of the color space, you can´t go beyond. Remember this is just an academic excersise</p>

 

<blockquote>

<p>IOW how did over saturating to the point of clipping give you something to measure in a tiff image to get an accurate 3D gamut plot? What was in the image?</p>

</blockquote>

<p>The plot of a gamut just consider the boundaries or most saturated colors, the content of the image itself is irrelevant</p>

<p>Following is a gamut plot of a test image (solid) compared to AdobeRGB (wireframe)</p>

<div></div>

<p>Here is the test image itself. The yellows that are outside the AdobeRGB are not from the colorchecker itself, but from the yellow marker. (This image has been converted to sRGB for web)</p><div></div>

<p>Here one more test, I took a colorchecker image to the extremes (saturation) and processed in DxO using Prophoto as output space and then another using Adobe RGB as output. Opening Both in Photoshop and converting the image in AdobeRGB to ProPhotoRGB (which in practical terms is pointless), gives the same values for every patch.</p>

<p>Then I plotted the gamut of the file that was encoded with ProphotoRGB against ProPhotoRGB and Against AdobeRGB and it basically fills entirely the AdobeRGB color space, due to the exagerated saturation, and there is nothing outside it. For me it is sufficient proof that DxO uses AdobeRGB when outputing TIFFs and there is no point in using ProPhotoRGB as output space.</p>

<p>If you want / need ProPhotoRGB from DxO, the only way I know of, is using DNG output and continue you editing in LR / ACR.</p>

<p>Following are the super saturated image in DxO and the gamut plots comparing the file encoded in ProPhotoRGB against ProPhotoRGB and then against AdobeRGB</p>

<div></div>

<p>Here is the plot of the image gamut (solid) compared to ProphotoRGB (wireframe)</p><div></div>

<p>Image gamut (solid) vs Adobe RGB (wireframe). They are basically the same volume</p><div></div>

<p>Hello Again,<br>

This discussion has evolved somewhat above my knowedge level, but based on Francisco's argument it would seem that in fact the only way to get an image out of DxO with a ProPhoto ARGB colour space is to export a DNG and not a TIFF. Is that correct?</p>

 

<p>David,</p>

<p>Yes, that has been my experience. I'll be glad if somebody else proves otherwise.</p>

<p>Anyway, keep in mind that a linear DNG (as the ones DxO Output) are useful ONLY if you continue your editing in LR or ACR. I don't know of any other software that handles those files.</p>

<p>As a side note, playing around with gamut plots, in my experience the majority of images fall inside the gamut of AdobeRGB. It is only with very saturated colors (i.e. flowers) or when you deliberately increase saturation that you run into colors outside AdobeRGB. No actual monitors can show them and only some high-end printers can output color outside AdobeRGB in the yellow-orange region.</p>