XRite ColorChecker Passport

[don_bryant2]

<p>Just curious if anyone has tried the Xrite ColorChecker Passport yet? It looks like an interesting product and has a reasonable price. I'm tempted to order one.</p>

[patricklavoie]

<p>i will be in Toronto next week, Vistek have it..im also curious ; )</p>

<p>Im using a nice small gray card up until now..and that is pretty usefull.</p>

[howard_m]

<p>$130 seems kinda steep. I was hoping for more like $99</p>

[digitaldog]

<p>Did beta. Having a color checker in such a small and protected package is useful. It makes very good DNG profiles (I compared them to the Adobe DNG profile editor for a 5DMII , the X-Rite profiles are a tad better). The warm and cool patches are again, in some rare instances useful for white balancing in a Raw converter. Those targets, which are hand made are expensive. But considering what a full sized and the newer mini checker now cost, seems like a deal considering all you get. </p>

[ben_goren]

<p>Donald,</p>

 

<p>I’m starting to come to the conclusion that there are three basic scenarios that make

sense, and I don’t think the new ColorChecker fits into any of them better than anything that already exists.</p>

 

<p>First, if you’re doing true color-critical work, such as product photography or art

reproduction, your <em>only</em> option is to create a profile for each combination of camera /

lens / lighting. And that will require photographing a multiple-hundred-patch target, building an ICC

profile, and applying said profile to the image. I like the Christopher Metairie Digital Target for use

with Argyll CMS, but Xrite has more user-friendly (and <strong>much</strong> more expensive!)

options.</p>

 

<p>Most of the rest of photography that needs reasonably accurate color really just needs a decent

white balance. And, to be honest, a $10 Kodak Gray Card is at least 95% as good as the most

expensive alternatives out there, if not every bit as good. Sure, some may have a slightly flatter spectral response…but the cameras themselves have a bigger variation in their own response curves, so it’s really not worth worrying about in real life.</p>

 

<p>Lastly are creative types of photography where you’re intentionally altering color for a

particular purpose. And while it might be nice to start with a decent white balance, in the end it really

doesn’t matter.</p>

 

<p>Cheers,</p>

 

<p>b&</p>

[digitaldog]

<blockquote>

<p>First, if you’re doing true color-critical work, such as product photography or art reproduction, your <em>only</em> option is to create a profile for each combination of camera / lens / lighting. And that will require photographing a multiple-hundred-patch target, building an ICC profile....</p>

 

</blockquote>

<p>I have zero need for ICC input profiles for digital cameras. Years of work have resulted in iffy benefits expect in very rare cases (totally controlled studio work). I do have a need for DNG profiles which this product produces and a ColorChecker is required for this or the free solution from Adobe. I have found the ColorChecker to also be a very useful color reference for some situations. </p>

<p>Those $10 gray cards are not gray, not spectrally neutral, at least those I’ve actually measured using a Spectrophotometer. </p>

[ben_goren]

<p>Andrew,</p>

 

<p>Just to make certain I’m not nuts, I re-measured my years-old $10 Kodak gray card with

an i1 Pro. D50 Lab values are 49.438321 -1.042986 1.088057, which is “good enough” for

non-colorimetric work. Attached is a graph of the spectral response which, again, isn’t perfectly

flat but is still plenty “good enough.”</p>

 

<p>Perhaps yours was an $8 gray card…?</p>

 

<p>Cheers,</p>

 

<p>b&</p><div></div>

[digitaldog]

<blockquote>

<p>D50 Lab values are 49.438321 -1.042986 1.088057, which is “good enough” for non-colorimetric work.</p>

 

</blockquote>

<p>The aStar and bStar of my X-Rite target read -0.1, 0.2, a heck of a lot better, more neutral than your -1.0/1.0 $9 Kodak card. More neutral by a significant factor. You can argue both are adequate but there’s no question the X-Rite target is far more spectrally neutral. </p>

[joseph_wisniewski]

<p>Ben, you're one of the first people I've seen use "like" and "Christophe Metairie" in the same sentence. His charts, despite the insane number of patches (or more likely, because of the insane number of patches) are done with process colors (ordinary CMYK printer or "press" colors), and produce basically three usable vectors (areas of the color space) to measure. The Macbeth ColorChecker gives you 18 patches that are <strong>not</strong> process colors, each made from a unique paint that gives you a spectral response much closer to that of real objects, so it can "fix" color much better.</p>

<p>So, a ColorChecker green reacts to changing light like leaves do, the ColorChecker skin reacts to changing light like skin does, while the Metairie charts react like pictures of leaves or skin would react.</p>

<p>Imagine cutting a color wheel up into three big slices of "pie", and dragging the hue and saturation of the piece and the green piece to make red flowers and lipstick and green leaves and makeup look better (but not as much better as the "better" paints on the ColorChecker green and red patches would do). But since we've only adjusted green and red, making those adjustments pulls the yellow/orange of fleshtones (in between red and green on the color wheel) off in a random direction, making the flesh tones look worse than they did to start. Now picture having 18 smaller pieces of pie, so the red flower and lipstick piece, the green leaves and makeup piece, and the red/orange skin piece all get fixed properly, along with the sky blue piece, and 14 other pieces....<br>

<br /> Charts like Metairie's are used for two purposes only.</p>

<ol>

<li>If the CMYK characteristics of the chart are know, they can be used to calibrate a device designed specifically for the processing of CMYK images, such as a scanner used to scan photographs or pictures in books. Like the classic IT-8 scanner calibration chart. </li>

<li>If they're printed on a device with a highly non-linear response (like an inkjet printer) they can be used to create printer profiles to compensate for the way dots of ink interact with paper and with each other. That's what Metairie's chart looks like, a printer calibration chart.</li>

</ol>

<p>A camera is "good enough" at color so once it's been calibrated with a chart that "means something", like the ColorChecker, it could then be used with Metairie's chart to calibrate the printer that he used to print the chart. But none of us really care about calibrating Christophe Metairie's printer for him.</p>

<p>Andy's right about ICC profiles and cameras. Those made from charts that have lots of patches but only three actual inks are near useless. I believe PictoColor InCamera has an option to profile a camera from an IT-8 chart. Profiles generated properly from a Macbeth ColorChecker, with 18 decent "divisions" of color space work a lot better. But even then, it takes really good software to make such a profile. Doesn't really matter if it's applied as a camera profile, Adobe style, or an ICC profile farther down the process, as long as it's a good profile. I've never seen an ICC profile as good as what the DNG profile editor puts out. If the X-Rite profiles are even better, it's a definite "good thing".<br>

Wizfaq</p>

[digitaldog]

<blockquote>

<p>Profiles generated properly from a Macbeth ColorChecker, with 18 decent "divisions" of color space work a lot better. But even then, it takes really good software to make such a profile. </p>

 

</blockquote>

<p>Right. Of course the “best” scenario would be to measure 16.7 million colors. The resulting profile would be larger than many images not to mention how long it would take to measure the data. What’s interesting is how some (X-Rite specifically) are doing some really good work at building profiles with so little data and doing a lot of extrapolation. In the very old days, I recall a product called Profile 80 which supposedly made output profiles with as few as 80 patches. And with some really well behaved devices, the profile was OK. Otherwise, they sucked! But this was back in the early 90’s. <br>

ColorMunki builds an impressively good profile with only 100 patches. You can then read more and iterate a new profile. That’s something I’ve yet to really need but it is interesting to see the technology evolve. </p>

 

<blockquote>

<p>If the X-Rite profiles are even better, it's a definite "good thing".</p>

</blockquote>

<p>In terms of the one camera I tested and the images I compared, its a bit better. But a lot of this is subjective. Especially useful is the tonal mapping of the X-Rite profile which are IMHO visually superior to DNG profile. Here’s a sample of a high ISO shot from the last Photoshop World Keynote (the guy in the jersey is John Nack, Product Manager for Adobe and his boss). Might be difficult to see on the web, but using identical rendering settings, the X-Rite profile seems to have more highlight detail:<br>

<img src="http://digitaldog.net/files/AlphaProfile.jpg" alt="" /></p>

[joseph_wisniewski]

<p>I just checked all four sets of instructions on Christophe's site, for using his chart with Adobe DNG Profile Editor, DxO, Hassleblad Flexcolor, and P1C1.</p>

<p>Am I totally misreading this, or in each of these cases is he using one square of the 570 patch chart to set white balance, and discarding the other 569 squares? None of these programs can profile with arbitrary charts, and as far as I know, none has any way of loading a reference file to correct this situation. I'm sure that's the situation for DNG Profile Generator.</p>

<p>Reports from people who tried to set the thing up for real profiling software (like Argyll) report worse results than simpler targets, like a classic IT-8. I'm not getting the point.</p>

[ben_goren]

<p>Andrew, I didn’t claim that the Kodak card was better; merely that it’s

“good enough” for most purposes, including basically everything where colorimetric

accuracy is <em>not</em> required.</p><p>Joseph, the first thing to remember is that the spectral characteristics of a chart aren’t

nearly as important for profiling an image captured with a camera as that the chart stimulates a response in human eyes —

or devices such as cameras that mimic them — that covers a large gamut with reasonably

even spacing. It’s especially the case if the target incorporates a similar gamut to the device

you’re eventually targeting for output. I’ve been using Christopher’s target

with studio strobes (sunlight in one case), and make a new profile (with Argyll) for every scene change (after using dcraw

to develop the RAW images into linear TIFFs). Perhaps not the most efficient workflow, but

it’s producing good results for the low-volume stuff I’m doing that actually needs

colorimetric accuracy. That said, I can tell you from experience that there’s not much point in trying to create a one-size-fits-all profile with the target.</p><p>Also Joseph, in response to your later post, I have no clue what Christopher does to build his profiles. Aside from buying the chart,

I’ve had no contact with him. As I mentioned, I’m using Argyll with good results (for

me, at least). A typical profile will have an average ΔE < 1, with the max usually in the mid-single-digits. In my workflow, the particular white balance doesn’t matter, so long as you use

the same for both images (with and without the chart).</p><p>I’ve just discovered DNG profiles. I fear I’ve got some long weekends / nights

ahead of me to build some. Even if I still use Argyll for the color-critical stuff, it’d be nice to

have a closer colorimetric starting point for the “artsy” things. If the oranges are to be

over-saturated, I want it to because <em>I</em> decided them to be that way, not because some unknown

code monkey somewhere thought they look better that way. Might even make it worthwhile re-visiting a couple favorite shots that I really struggled with in post.</p><p>Cheers,</p><p>b&</p>

[ben_goren]

<p>Andrew,</p>

 

<p>I gotta say…the Xrite may be marginally better, but only on very close examination.

It’s put a bit too much magenta in the skin tones, whereas the Adobe version has too much

green. And in both, the highlights are severely blown and the shadows hopelessly blocked. Maybe

that’t the light the room was lit with, but regardless…this isn’t a photo where

I’d be worrying about colorimetric accuracy. And I guarantee you you’re much more than

just a couple ΔE off. What good is your 0.1 ΔE gray card when the results are as colorized as either version is?</p>

 

<p>Cheers,</p>

 

<p>b&</p>

[digitaldog]

<blockquote>

<p>I didn’t claim that the Kodak card was better; merely that it’s “good enough” for most purposes, including basically everything where colorimetric accuracy is <em>not</em> required.</p>

 

</blockquote>

<p>Perhaps. I’m not going to say for all users here that its “good enough” because I have no idea what their level of acceptance is. Point is, the neutrality of the X-Rite product is significantly better, and in addition, there’s a lot more available in the Passport for other tasks that warrant, for me, the difference in price. </p>

<p>There’s a significant difference in DNG camera profiles and ICC camera profiles. One if major reasons Adobe doesn’t support ICC profiles is they are by their very nature, output referred, the DNG profiles are scene referred. Eric Chan of Adobe posted this on the ColorSync list which I think is a good start in defining the differences:</p>

<p>Camera Raw's color profiles are all based on scene-referred colorimetry and deliberately ignore tone curves (i.e., since everything is linear in the raw stage, any required exposure adjustment is just a simple scaling of the raw data by a constant). The idea being that the profile sets up the initial scene-referred color and it's then up to the user to guide the rendering process, mostly through the use of a non-linear tone mapping curve (looks like a big gamma adjustment with an S-curve, by default). So if you're evaluating the final output, which it seems like you're doing, it's not at all surprising that you're seeing large mismatches, esp. if you're including L* differences in your error metric.</p>

[ben_goren]

<p>Andrew wrote:</p>

 

<blockquote><p>Perhaps. I’m not going to say for all users here that its “good enough” because I

have no idea what their level of acceptance is.</p></blockquote>

 

<p>You don’t need to. Unless you’re doing precise colorimetric rendering —

however it is you do that — you’re going to wind up with variances of at least several

ΔE in all directions over the entire spectrum. Compared to that, the Kodak grey card is,

spectrally, a perfectly flat line.</p>

 

<p>It’s like using a micrometer in wood-frame housing construction. Sure, it’s more

precise — but it gets so lost in the noise it’s meaningless. Machinists need

micrometers, not carpenters. Even cabinetmakers don’t need micrometers.</p>

 

<p>Which brings us back full circle. If the Passport works for you, great, but to me it sure seems

like it’s straddling a no-man’s land, where you’d either need a lot more patches

to get a truly accurate colorimetric rendition of a carefully-controlled scene, or the work

you’re doing is such that nobody’ll notice a few ΔE here and there any more

than they’ll notice that their ceilings are “only” 7′

11^{<small>119</small>}⁄_{<small>128</small>}″ high.</p>

 

<p>Cheers,</p>

 

<p>b&</p>

[digitaldog]

<blockquote>

<p>Which brings us back full circle. If the Passport works for you, great, but to me it sure seems like it’s straddling a no-man’s land, where you’d either need a lot more patches to get a truly accurate colorimetric rendition of a carefully-controlled scene, or the work you’re doing is such that nobody’ll notice a few ΔE here and there any more than they’ll notice that their ceilings are “only” 7′ 11^{<small>119</small>}⁄_{<small>128</small>}″ high.</p>

 

</blockquote>

<p>IF all the Passport did was provide WB functionality (which for Raw capture is ideal, NOT a Kodak Gray card), I’d agree with you. Obviously it provides far more functionality but at a price point. </p>

<p>IF you don’t have a ColorChecker, its a good deal and useful. If you do, then its questionable if this is a good purchase. For location work, the mini CC is nice and portable like the Passport. But no protection and not easy to shoot without either an assistant or someone holding it. The Passport’s hinged tri-door design makes it much easier to place the product in a scene without much effort. You also get the off colored WB squares for use WB targeting in a Raw converter. You could just as easily move the Tint and Temp sliders to get a desired rendering, but the slider is pretty course in movement and you’ll get subtler results using the patches. <br>

There’s no comparison between a $10 Kodak gray card and this product! One is not appropriate for White Balancing Raw linear encoded files and its a one trick pony. Again, if don’t have a ColorChecker, this is a nice addition to the camera bag, especially if you want to build DNG profiles. <br>

I’m hard pressed to believe, after years of testing virtually every ICC camera profile generator on the planet that its going to provide anything close to a colorimetric match to the scene, especially considering they are output referred! So I want pleasing color (colorimetrically correct, scene referred color is butt ugly**). So the question becomes, what tool provides a good starting point for final, output referred rendering. Could be an ICC profile but that’s not an option in Adobe converters. That leaves us with DNG profiles. So considering that we’re not going to get colorimetrically correct, scene referred imagery, that the converter’s goal isn’t to provide that, it seems moot that the profiles use 24 patches or 2400 patches. I’ve yet to see where a custom DNG profile, either generated with Adobe or X-Rite’s software don’t produce more preferable rendering by simply loading said profile compared to the profiles that are shipped in ACR/LR. So if you’re using an Adobe converter, just get yourself a Macbeth (or Passport) and stop worrying about the number of patches and make attractive, output referred, non colorimetrically “correct” images. </p>

<p>**http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf</p>

[ben_goren]

<p>Andrew wrote:</p>

 

<blockquote><p>I want pleasing color (colorimetrically correct, scene referred color is butt

ugly**).</p></blockquote>

 

<p>That’s fine. And, for most of my photography, that’s what I want, too.</p>

 

<p>So why are you getting all hung up about fractional ΔE precision, if you’re just

going to then proceed to radically alter the colors? It’s like insisting that the lumber yard sell

you two-by-fours that are 6′ long

±^{<small>1</small>}⁄_{<small>64</small>}″, when all

you’re going to do is eyeball them to about knee height and cut them with a hand saw.</p>

 

<p>When you <em>do</em> care whether your white balance target is ±1.0 or ±0.1

Lab units is when you’re doing exactly what you consider “butt ugly.” A painter

doesn’t want you photographing her work and getting back your interpretation of better

colors; she wants exactly the colors she used — even if they <em>are</em> “butt

ugly.” And <em>that’s</em> when it’ll matter how flat the spectrum of your gray card

is. (But it really won’t, because you won’t be depending on a mere gray card to get

colorimetrically-correct color; instead, you’ll be doing that other thing you hate with creating a

custom ICC profile for just that one picture.)</p>

 

<p>After all, super-high precision is meaningless without correspondingly high accuracy. Unless you’re the type to use an atomic clock to figure out how long to roast a bird?</p>

 

<p>Again, to bring it back to the topic: the Passport is a $50 kitchen timer with a display that reads in hundredths of a second. A total waste of precision and money, but it’ll still give you an idea of about when your goose is cooked.</p>

 

<p>Cheers,</p>

 

<p>b&</p>

[digitaldog]

<blockquote>

<p>So why are you getting all hung up about fractional ΔE precision, if you’re just going to then proceed to radically alter the colors? </p>

 

</blockquote>

<p>I’m not hung up, I’m pointing out that you are getting a vastly more spectrally neutral white balance target for the money! And when I WB, I want to do so on something that’s neutral which does affect the rendering which isn’t necessarily colorimetrically correct and for ETTR, I want something with a very high Lstar value as well. </p>

 

<blockquote>

<p>Again, to bring it back to the topic: the Passport is a $50 kitchen timer with a display that reads in hundredths of a second.</p>

 

</blockquote>

<p>I disagree and have pointed out where its a very good value for those people who have no such targets. I’ve also worked with this product for half a year, how long have you to base such an opinion? The price point is what it is, and given the precision of the targets, so much the better. And again, a $10 gray card is absolutely the wrong tool for Raw processing WB, whether the a and b star values are ±1.0 or ±0.1. The company states a certain degree of spectral neutrally and delivers that. </p>

[mike_curtis3]

<p>I just purchased the Passport and played with it last night under various lighting conditions. I am impressed with the DNG profiles I created so far looking at the results on my calibrated monitor. I am anxious to play with the Creative targets with respect to landscapes and portraits.</p>

<p>Mike</p>

[roger_smith4]

<p>Thanks for your thoughts, Andrew. Interesting that you find any custom DNG profiles better than what Adobe ships. Why do you think they don't do a better job out of the box?</p>

[digitaldog]

<p>I would highly suspect a custom DNG profile is "better" for the same reason all custom profiles have the potential to be better; you're building a custom response to your camera chip in this case. </p>

[Tim_Lookingbill]

<p>Did anyone catch Ben Goren's mentioning his shooting artwork? Accuracy is going to be hard to achieve shooting this type of scene using the a DNG CCchart profile because artwork can contain pigments and dyes whose spectral response is completely different from the pigments used in the X-rite Color Checker.</p>

<p>When I take photos of inkjet and CMYK prints to demonstrate the appearance of color tint cast between viewing lights like the Solux and the GE Sunshine, I can't use my DNG CC chart profile. I have to use Adobe's supplied canned ACR 4.4 profile or else I have to do some severe hue/saturation adjusts to make the image appear as my eyes see the scene which is right next to my display so I know when its off since I'm not relying on my memory of the colors.</p>

<p>Does this new Passport address these spectral differences to attaining accuracy over pleasant colors? Did X-rite change the formula for the pigments used in the patches?</p>

[digitaldog]

<blockquote>

<p>Accuracy is going to be hard to achieve shooting this type of scene using the a DNG CCchart profile because artwork can contain pigments and dyes whose spectral response is completely different from the pigments used in the X-rite Color Checker.</p>

 

</blockquote>

<p>This is not unique to DNG profiles or for that matter any capture profile. And the only way to properly define accuracy is using measured colorimetric values of the original, by the time you have an image you can view or print, its output referred so the term accuracy files out the window:http://www.color.org/ICC_white_paper_20_Digital_photography_color_management_basics.pdf</p>

[Tim_Lookingbill]

<p>Andrew, it's reproduction of artwork we're talking about here. </p>

<p>Artists don't care what the accuracy as defined by colormetric representation of the numbers of the image is. They just want it to look like their paintings and I've found that using the CC chart DNG PE profile creates more work than it's worth when trying to achieve this.</p>

<p>How do you explain that inaccuracy?</p>

[digitaldog]

<p><br />>>How do you explain that inaccuracy?<br /><br />No, I explain it as a visual match. <br /><br />Accuracy by its definition requires some matrix that can be used to signify that something is either accurate or not accurate and by a fixed value (in color, that's deltaE). In color, we use colorimetric measurements to do this. The very fact that two colors can appear to perfectly match yet measure differently, something useful we call metamerism or a metameric match is quite useful but can't be defined as a accurate measured match because the values are so different. <br /><br />Accuracy in this context is often used incorrectly, or in an undefined way, or as a marketing buzz word. That we result in a match is good enough, we don't have to label this incorrectly as accurate, especially when there isn't a lick of measured values to express accuracy. <br /><br />Accuracy (see #2):<br />1 the quality of nearness to the truth or the true value; "he was beginning to doubt the accuracy of his compass";<br />"the lawyer questioned the truth of my account" <br />2 (mathematics) the number of significant figures given in a number; "the atomic clock enabled scientists to measure time with much greater accuracy"<br /><br /><br />>>Artists don't care what the accuracy as defined by colormetric representation of the numbers of the image is.<br /><br />Because that's usually not possible. If I provided you a colorimetric, accurate set of values, in all likelihood, the result would be a visual mismatch. So accuracy isn't at all a useful term here, in fact its simply confusing and muddies the conversation.</p>