Color Calibration Comparison: Canon 5D Mark lll from Different Profiles

In my recent article comparing the color response of the Canon 5D Mark lll in Lightroom 4.1 with and without SpyderCheckr calibration received a lot of interesting responses. The most frequent questions were: what profile did you use (answer: Adobe Standard, for reasons that will become apparent below), and what effect would changing that profile have on the color? That seemed like a valid line of enquiry. I also thought it appropriate to also show what would happen if I build a calibration on top of each of these profiles, and applied these calibrations for comparison. The usual caveats about what you see on your (possibly uncalibrated) monitor through your (possibly non-color managed) browser apply.

First its necessary to understand the importance of defaults in such settings as Tone Curves and Profiles. Images are imported into Lightroom, and unless  you build a special Preset, and specifically include things like a Tone Curve and a Profile into that preset, and remember to import all images through that Preset, then you (like 99% of Lightroom users) will end up processing your images on top of the default settings. So my standard recommendation is to leave the Tone Curve and the Profile at the default, and to build your custom SpyderCheckr camera calibration (and your SpyderCube white balance and exposure adjustments) on top of those settings. That way you won’t have to remember what settings were in place when you created the calibration, and the chance of you having the wrong settings in place when you apply the calibration are minimized.

Also note that pushing the Reset button in Lightroom’s Develop Mode will reset the Tone Curve in Process2012 to Linear, and the Profile to Adobe Standard, even though you might not have intended either of those actions. One more reason its safest to build on top of the defaults, to ensure consistency, and minimize errors.

Next, I built three different calibrations in the SpyderCheckr software, all from the same shot of the SpyderCheckr target, taken with the Canon 5D Mark lll camera. In all cases I left the Tone Curve set at the new default of Linear. In all cases I left the intent in the SpyderCheckr application set to Colorimetric, the most literal setting. For one calibration, I left the Lightroom 4.1 Profile set at the default profile of Adobe Standard. For another I set the Profile to Camera Standard. And, to test one of the “flavored” choices, I also build a calibration with the Profile set to Camera Portrait.

I then opened the same 5D Mark lll test image I used in my previous articles in Lightroom 4.1, at its default Profile setting of Adobe Standard, and captured what the color from that looked like. I then applied the Lightroom Preset I had build for the Mark lll at Adobe Standard, and captured that as well. These two images were combined side by side into an Uncalibrated/Calibrated image. That image is below. Click on it to see the results at 100%.

Results using Adobe Standard, Uncalibrated, then Calibrated

These results, though I created a new calibration, and newly applied it to the image, are similar to the results in the previous article, as the same configurations are used. Next I repeated the same test, with Lightrooom 4.1 now set to the Camera Standard Profile. The results for this pair of images are below.

Results using Camera Standard, Uncalibrated, then Calibrated

These results are quite similar to the first pair. Camera Standard is not quite as oversaturated to Adobe Standard, showing a bit better color detail since it is not as bleary, and with a somewhat different hue. Calibration, starting from these two different hues and levels of oversaturation, is satisfyingly similar. This is what calibration is supposed to do. Calibration takes two inaccurate and dissimilar monitors, and makes them both correct and similar by calibrating them. Camera calibration similarly corrects both versions of the image, and as a result makes them a closer match as well. This is very reassuring.

Now to test with a “flavored” Profile. One intended for Portraits is shown below.

Results using Camera Portrait, Uncalibrated, then Calibrated

Here the  reds are less oversaturated then either of the previous cases, but the near whites show much pinker and more saturated. Note that the calibrated result on the right adds a bit more color detail in the reds, and reduces the over saturation and excess density in the near whites considerably. As a calibration can’t fix everything, there are compromises required. The worse starting point of this Profile resulted in less accurate reds than the other calibrations, in the process of curing the heavily over saturated and too-dark near white zone. This result would indicate to me that one of the other Profiles would be a better starting point for calibration, much as it is possible to choose a better media setting in a printer driver before profiling the printer.

It makes sense that avoiding intentional flavorings and non-linear Profiles in advance of creating a Camera Calibration is a good idea. The fact that the default (Adobe Standard) Profile offers good results in convenient, as it allows users to feel comfortable doing what is easiest: calibrating on top of the default, for the least chance of later errors in applying the calibration to future images as they are imported.

Credits: C. David Tobie, Copyright 2012. Website: CDTobie.com Return to Blog’s Main Page

Advertisements

Color Calibration: Canon 5D Mark lll

Calibration Corrections for the 5D Mark lll

In my recent article comparing the color response of the Canon 5D Mark ll to the 5D Mark lll, I showed the calibration adjustments from Datacolor’s SpyderCheckr for both cameras. Today I am focussing on the Mark lll, and what calibration does to improve its color response, so I will limit myself to the Mark lll correction chart only.

Color Corrections for Canon 5D Mark lll from SpyderCheckr

While there are several colors where the combination of hue and saturation adjustments are fairly significant, I will limit my examples to the primary color channels with the largest Hue shift (Green, with Hue -16, and Saturation -4) and the largest Saturation Shift (Red, with Hue +6, and Saturation -15). The first question is: are these adjustments visible in real world images? And the second is: do they actually improve the results in images where they are visible? I selected images from my first test shoot with the 5D Mark lll that showed these adjustments clearly, and which can be used to judge the value of these calibration adjustments.

Limitations of Web Color

The images you will see below were color accurate on a calibrated display. On an uncalibrated screen, or with a non-color managed browser, the results may not be as accurate. However they should still show the degree of difference, even if the colors are not exactly as they would be under color management. So try to visualize what the text describes for color and saturation, and use the images to compare the degree of change, if the actual changes described are not reflected on your screen.

First the Reds

To judge the effect of red adjustments, I selected am image with saturated red flowers, and cropped a section at 100 percent. The red flower petal in the center of the image moves forward from the point of focus, and becomes less focussed in the lower regions. The reds are extremely saturated, and the trained eye sees a color that will be problematic to print as well as having the bleary quality that over saturated colors often exhibit.

5D Mark lll Reds, Uncalibrated

Next we’ll examine a similar closeup from the same image after the SpyderCheckr calibration has been applied to the image. Here the colors look more believable, and show a good deal more color detail, that was lost the the bleariness of the uncalibrated version. The Hue Shift, while far smaller then at the Saturation reduction, also improves the realism of the image, accurately reflecting the transition from magentas to a more orangey red in this type of flower. Its often surprising to see that color correction does not just improve the colors; it improves the color detail, which results in a more detailed image; something we tend to associate with focus and lens quality, when it can actually be an artifact of incorrect color.

5D Mark lll Reds, Calibrated

Now the Greens

Next we’ll move on to the greens, which are a subtler situation, since it is the Hue, not the Saturation, that has the largest adjustment. Here the issue is less one of lost detail, than one of lost color richness. Leaf greens must contain an appropriate yellow component to read correctly to the eye; an emerald green leaf looks as false as the Emerald City, painted onto the Wizard of Oz backdrop.

5D Mark lll Greens, Uncalibrated

Now we’ll compare this with the calibrated green, which has a 16 point Hue shift towards Yellow, and a 4 point Saturation reduction. Here the greens read as much more realistic, with the necessary yellow component, and the slight reduction in saturation to produce photo-realistic foliage. This will be a much more printable green, and much more satisfying in print, as well as the improvement shown on screen.

5D Mark lll Greens, Calibrated

And the Rest

Other calibration corrections for the 5D Mark lll follow a similar pattern, but the effect is reduced for those channels requiring less adjustment. Overall, native color on the 5D Mark lll is very good, but calibrated results are truly excellent.

Personal Preference 

The color corrections shown above are aimed at accurate color portrayal. They are from the Colorimetric option in SpyderCheckr. Personal preference may lean towards emerald green leaves and over saturated flowers; thats fine as artistic intent, but it should occur by intentional adjustment, not by incorrect original coloration.

A Note to Nikon Shooters

Vincent Versace recently requested that I do a color comparison of the latest Canon and Nikon models. As a Nikon shooter, he as always felt that the Canon results were more saturated; perhaps too saturated. The calibration corrections above do tend to lean towards desaturation of the RAW file’s color; in fact there is not a single channel where calibration increases the saturation. I plan to analyze the color of the new Nikon D4 against its predecessor, the D3, and similarly the D800 against the D700. Once I have completed those results, I will be in a position to compare the 5D’s color saturation, and other color characteristics, to those of the D4 and D800. Perhaps the results will reinforce Vincent’s personal impressions. It will be interesting to find out.

A Note on Camera Color

What cabbage leave does camera color come out from under? It has two parents: the camera manufacturer, who is responsible for the hardware, and for the firmware that produces the proprietary file, and the RAW converter developer, who uses their own form of camera profiles to convert that proprietary camera format to a RAW file on screen. At one time, both these were the same parent; and we felt obliged to use the manufacturer’s RAW converter when testing the camera. But today the majority of RAW files, at least those from DSLR and subDSLR cameras, are processed in third party software. And by far the most common of those third party products is the Adobe engine used in Lightroom, ACR, and therefore Photoshop. So when I talk about camera color, I’m speaking of the camera color we are going to get in the real world, where we process our images in Lightroom or ACR. Using another third party RAW converter, or a camera manufacturer’s conversion utility, may produce different results. But given the success I have been having with the Adobe engine and a SpyderCheckr HSL correction set, I am quite comfortable with the color from this workflow, and don’t see any significant reason to be looking at less common or more complex workflows in order to achieve the results I need; especially with Lightroom 4’s much improved capabilities.

Credits: C. David Tobie, Copyright 2012. Website: CDTobie.com Return to Blog’s Main Page

Lighting a Color Target

Its not rocket science; but the lighting of a color target requires a clear understandings of the functions, and the light sources, involved. This article will break down the various factors. The ideal way to light a target is with a single point light source at a great distance, at 45 degrees to the face of the target, with no other lighting. The rest is all details…

Why 45 degrees? Because the ideal response from the target patches is the “45/0” response, just like a spectrophotometer or colorimeter that illuminates the sample with light from 45 degrees, and measures the results from zero degrees, meaning perpendicular to the surface. There are other reasons this is ideal, including minimizing gloss reflections, offering a consistent balance between texture shadow and color, and avoiding complications that come with either straight-on lighting, such as an in-camera flash, or raking lighting, which accentuates shadows and unevenness.

Why a point light? Simply to make the light source as consistent and uniform as possible. Adding a reflector to a light that is very far away won’t cause problems, but simply moving it closer (if it does not offer enough light at that distance) and not adding the reflector would be a better choice.

Why a single light source? Its not impossible to light a target evenly with two lights, say from 45 degree angles to both sides, but its highly improbable that this method will actually provide even lighting. One theory of double lighting is that each light “falls off” a bit towards the farther edge of the target, and the second light, falling off in the opposite direction, will balance this out. The better way to avoid fall off, is to use a more distant source, such that the fall off is negligible. Secondary light sources are really about reducing shadows and harshness in images, and the image of a target is not intended to be pretty, its intended to be literal. Harsh is good, in this situation.

Other factors requiring mention here are the target and the camera, since it is the relationship of these two items with the light source creates the triangle that makes this all work. First, the target needs to be stably placed, perpendicular in both axes from the camera (having a tripod mount built into the target simplifies this). And the camera needs to be stably placed in relation to the target. Two tripods are the ideal solution, or a tripod for the camera and a light stand for the target. But with care, its possible to shoot an effective target image without either of these aids.

Now a bit on the “why” to go with the “how” above. Targets are shot to represent multiple things, or the workflow that the combination of these items creates. The light source is one factor, though a good “low metamerism” pigment target, once corrected for whitepoint and exposure, will offer very similar results under different light sources, assuming they are all fairly “good” (meaning smooth across the spectrum) lights. If you are shooting a particularly important set of images (say the entire clothing collection of an important commercial client) and want to be extra sure that you have the most accurate color results, a specific correction for this set might be justified, but in most cases, unless the light source is particularly difficult, a unique profile for the light source is unnecessary; all that is required is white balance and exposure data for the scene, which could be gathered from a color target, though other devices (particularly the Datacolor SpyderCube) are better suited for that job.

The next factor is the camera, by which we mean this particular camera unit; not the model line, but the physical camera in your hands. A generic camera color profile for a model line might be better than no correction at all, but it won’t address the unique issues of your particular camera, which is why custom profiling is important here.

If the camera’s lenses are interchangeable, and if you find the color is not quite the same amongst the lenses you own, then that is a final factor, one that may force you to create multiple profiles, for each lens, rather than a single one for the camera.

Now a bit on real world camera profiling. The photo below shows a perfectly acceptable target shot. There are no tripods involved, but the geometry is sufficiently accurate to do the job. The target is perpendicular, in both directions, to the camera, so there is little distortion of the target’s rectangle.

The light is single source, and comes from about a 45 degree angle to the target face. The light source is not a point, but its limited to a single patio door, with the hotel room blackout curtains closing off all but that one door’s glass. And the light source is a very spectrally smooth light source: the sky. While skylight is very blue (as high as 9300k), it is very even, and once white balanced and exposure corrected, this target capture will offer nearly identical patch values to one shot with a very warm but full spectrum light source, such as a tungsten lamp.

This particular target capture was taken because we were about to take a series of shots of artwork on this same coffee table, and needed reliable color in the resulting images, on a laptop which did not then have a color profile for the camera. This target capture did an excellent job of providing camera color correction for this series of images, and in conjunction with calibrating the laptop monitor, gave us sufficient confidence in our color to allow the entire project to be shot and processed on site.

Credits: C. David Tobie, Copyright 2012.   Website: CDTobie.com   Return to Blog’s Main Page