Procedure for profiling a digital camera
Notice: As of February 2007 we no longer offer a matte target. After years of testing, we have determined that the accuracy of the glossy IT8 target is superior to the matte target, making the IT8 target a better choice for profiling all devices. In addition, with careful preparation, the glossy IT8 target can be used to profile cameras with no issues related to reflections or glare, so a matte surface is not necessary. The following is taken from one of Mike Chaney's Tech Corner articles at Steve's Digicams and it describes how to profile a camera using the included IT8 target and Profile Prism.
I. Profiling a Camera with an IT8 Target
I am often asked about camera profiling in one context or another, and even challenged by other professionals as to whether or not it is even possible to develop ICC profiles for digital cameras. As I often say, the answer can be complex and may depend on many factors, but let's break it down into a few key points that are relatively easy to understand.
Profiling a camera: the process
In the early days of digital cameras, it was possible to produce a profile for cameras shooting in JPEG/TIFF mode mainly due to the fact that some cameras produced gross errors that could benefit from correction, even if the result wasn't completely "accurate". Now, most cameras comply reasonably well with the sRGB color space and many more advanced cameras even offer an option of sRGB or Adobe RGB as the color space used by the camera. When we have a relatively recent camera model and/or a color space selection, it is rarely beneficial to try to develop ICC profiles for the camera shooting in JPEG/TIFF mode because it is difficult to impossible to produce corrections that result in any consistent improvement. If we shoot in raw mode, however, most raw conversion tools offer an option to turn off color management so that custom ICC profiles can be created/used. With color management turned off, the raw data offers a much more consistent starting point, and profiling becomes not only possible, but often quite beneficial.
The process, at least conceptually, is very simple. Take a shot of a color target in raw mode, develop the raw image with color management turned off in the developing software, and use a profiling tool to create a profile from the image of the target. The profile can then be activated in the raw developing tool. That said, the actual process itself can get a bit complex if we want to ensure a quality profile. You need to get a good shot of the target under good lighting, and you need to use a profiling tool like Profile Prism that was designed with camera profiling in mind as camera profiling requires specialized options like the ability to normalize tone curves and let the device dictate white balance. There are other high-end (read expensive) tools that allow you to develop camera profiles. These tools offer specialized targets and software, but I find that with some care, it is possible to match or even exceed the performance of these "high dollar" tools with Profile Prism and a standard IT8 target!
The problem, the solution
Before we start with the details, it is appropriate to inject a bit of reality here. Many raw developing tools, while they are designed to produce the best color possible, just weren't built using any real "scientific" means for color accuracy. Some use a simple color matrix to tweak color so that it looks acceptable and many don't employ reasonable tone curves to ensure good shadow detail. In layman's terms, this is the reason that it is often possible to develop ICC profiles for raw images that result in better color reproduction than the raw tools offer out-of-the-box.
If we can develop a profile that improves color over the "default" color reproduction of the raw developing tool, we can say we have a successful/useful profile. Some may question whether or not it is possible to develop a single profile that works under all lighting conditions, or whether it is imperative to develop one profile for each lighting condition: sunlight, fluorescent, incandescent, mercury vapor, etc.. Again, the true scientific answer here can get complex, but I've found that when profiling the true raw data, a "generic" profile can be developed using direct sunlight. As lighting conditions (color temperature) shift from direct sunlight to warmer lighting such as incandescent lighting, the profile will become less accurate but the shift is not normally so extreme as to cause gross errors. This is, in part, because the color filters used on the image sensor aren't changing under different lighting. Their overall response is the same under different lighting and color temperature only affects the proportions of red, green, and blue recorded by the sensor. A good profiling tool can discover the overall color characteristics of the sensor which tend to be valid over a wide range of lighting conditions. Here, the closer you can get to the actual raw data the better, because up-front color corrections only tend to multiply color shifts, so a raw tool that offers the ability to process the raw data without injecting color corrections will work best.
While some may choose to develop different profiles for different lighting, and that's certainly optimal, a generic profile for sunlight should work under a variety of conditions. Shooting the IT8 target in direct sunlight helps to reduce any metamerism of colors on the target and ensures a good match to the data file that tells the profiling software what the color on the target should look like. Shooting in direct sunlight also offers the ability to eliminate glare as the IT8 target is a glossy target that, when not shot under the proper conditions, can certainly produce glare which will make the profile useless. Shooting with the light hitting the target at an angle is imperative to eliminate glare/reflections and due to the fact that our light source (the sun) is so far from the target, we don't have to worry about the light being brighter on the side of the target closest to the sun as we would with angled studio lighting! Here's how to shoot an IT8 with no reflections or glare:
Of course, a lot depends on your location and the time of year, but in general, the best time to shoot the target is either 1-2 hours before mid-day or 1-2 hours after mid-day. Try to shoot on a day with minimal clouds so the sun isn't changing intensity/color as you shoot.
It is helpful to attach your IT8 target to a piece of thick cardboard using small tacks or pins at the corners or even tape at the corners as an IT8 will tend to curl and bend when it heats up in sunlight.
Try to find a room where light is entering a window/door at a sharp angle and hitting a wall adjacent to the window. If you can open the window to reduce lighting variations caused by the glass, all the better! Here in the northern hemisphere, a south facing window often works well in the afternoon. If the sun doesn't hit a wall, a palette, chair, or other object may be used to place your cardboard w/IT8 in the sun.
Make sure the room is as dark as possible and that the only light entering the room is coming from the window. Also try to avoid the direct sunlight hitting bright colored (non-neutral) surfaces such as red walls, blue floor tiles, etc. as these reflections can cause color shifts on the target.
Place your target in the sunlight so that the sun is hitting the target at an angle and you can sit in the shadows while taking the shot. The following is a typical setup for shooting an IT8 target in direct sunlight. Notice how the sun hits the target at a sharp angle so that the camera can sit in the shadows, thereby eliminating glare on the target:
If your camera has a custom white balance feature, using a white/gray card or a white sheet of copy paper (don't use photo paper with brighteners), place the card at about the same location as the IT8 and make sure it is in the sunlight. Use the custom white balance on your camera to white balance on the card.
Take several shots of the target in raw mode. Take one "normal" shot and then increase exposure incrementally, taking several more shots with brighter exposures making sure to stop just before the exposure gets "blown out" in the highlights. Camera settings like aperture usually have little influence, but smaller apertures often produce more even lighting across the frame. Note that camera lens and ISO speed can make a slight difference in profiling, so be sure your ISO speed is set appropriately and you are using your most-often-used lens. If you and/or the camera are sitting in the shadows of the room, you can take the photo straight-on at the target and you should get no glare or reflections. When taking the photos, fill only about 3/4 of the frame with the target. Don't zoom in so far that the target covers the entire frame because light falloff from the edges of the lens can be a factor here.
Once you have the shots of the target, turn off color management in your raw developing tool and develop the photos. Depending on the raw tool you are using, turning off color management may entail selecting a color management tab and selecting "Embed camera profile", or selecting "None" in the "color management" dropdown. Whatever you do, the important thing to remember is that you need to be able to turn off color management to develop the profile. Then, once you are done creating the profile, the profile can be activated in the raw tool by selecting the ICC profile that you created. Of course, this assumes that the raw tool you are using allows selection of custom profiles. Not all tools allow use/application of custom profiles so be sure the tool you are using has this feature. The more popular third party tools like Bibble, Capture One, and (the now discontinued) RawShooter allow the use of custom profiles. When developing the images, develop to TIFF (you can use 8 or 16 bit/channel TIFF format).
In Profile Prism, click "File", "Open" and open one of the developed images of the IT8 target. Next, make the following selections on the Profile Prism main window (description and file name are just an example):
|Type of device to profile||Camera/scanner|
|Reference target||Choose the file for your IT8 target|
|Profile description||Something like "Canon 5D Generic"|
|File name||Choose a name like canon-5d.icm|
|Profile for||Highest Accuracy|
|White balance||Device dictates WB|
|Tone reprod. curves||Gamma Match (Auto)|
|All other options||"Normal" or zero (0)|
The above parameters are appropriate for profiling a camera. Once you have set all the parameters, mark the 4 corners of the target on the image of the IT8 target. The step by step procedures for profiling a camera or scanner in the Profile Prism help will show you how and where to place the crop markers on the IT8 target. Once placed, there should be a white punch-out in each of the color squares on the IT8 including the gray scale at the bottom. If the punch-outs don't align inside each color square on the target, the corner markers have not been placed properly. Finally, click "Create Profile" at the bottom left and Profile Prism will create your camera profile. You can then test the profile by selecting the profile in your raw developing tool using the file name you used in the table above. Once the new custom profile has been set, simply redevelop the photos and evaluate them for color accuracy/appearance.
Since some raw tools like Capture One and RawShooter apply some "pre-curves", it isn't possible to profile based on truly raw data. As such, you may have to create a profile for each of the exposures (the one normal exposure and several brighter ones) and then pick the profile that has the tone curve (shadow and highlight detail) that you prefer. Usually, the best result occurs when the curves displayed in Profile Prism (after clicking "Create Profile") end as close as possible to the upper/right corner of the graph. If the curves end on the top edge or the right edge of the graph, you may need to try a different/better exposure. Note that it is best to pick a different shot with a different exposure as opposed to tweaking the exposure of a single shot in the raw developing tool! With a little practice, the above process can produce excellent profiles for any camera shooting in raw mode. The above are the procedures we used to develop our own camera profiles for numerous raw tools. These profiles have gotten many positive reviews and are often compared to profiles produced with much more expensive equipment/targets from other sources.
II. Using the Software
Step 1: Specify profile details
Before telling Profile Prism to generate your ICC profile, you need to specify a name, description and some options. Below is a description of each entry.
- Type of Device to Profile: Since we are profiling a digital camera in this section, select "Camera/Scanner" for this setting.
- Profile Description: Enter a description for the profile such as "Acme 800 camera profile - sunlight".
- File Name: To save your profile to the system color folder (c:\windows\system\color for example), click the "..." browse button and enter a new file name in the "Save As" dialog or select an existing profile to overwrite. You may also type a file name directly into the edit box if you like, however, if a file name is typed without specifying a path, the profile will be saved in the folder where Profile Prism is installed. For example, entering "test" in the file name field will result in a profile named "test.icm" being stored in your \program files\prism folder.
- Profile For: Profile Prism can create very accurate profiles, however, since some digital cameras can be inaccurate in very narrow bands of the color spectrum, profiles for these cameras can make images look noisy. This is due to the fact that you are taking a very narrow band of color and expanding it to have wider "separation". In doing so, shadow/background noise may be enhanced. Generally the more expensive "prosumer" cameras can be used in "Highest Accuracy" mode. If you find that your profiles create noisy images or images that have posterization artifacts, however, you may need to switch to "Smoother Gradients". The "Smoother Gradients" method sacrifices a (small) amount of accuracy to ensure that there are no large color variations. If you are unsure, always start with "Higher Accuracy" and only use "Smoother Gradients" if you have a problem with image noise.
- Tone Reproduction Curve: If you are profiling a camera or scanner, you may select from the following choices for how Profile Prism generates the tone curve for your profile:
- Normalize: The Normalize option is the default selection and should be used whenever you want to reproduce accurate (linear) tonality on a device like a camera or scanner set to automatic settings. On such devices, exposure is relative and often changes from one capture to the next, so normalizing the tone reproduction curves will adjust for any slight deviation in exposure of the test target in order to create curves that will work well with any shot regardless of how it is exposed. Note that when profiling some devices, particularly cameras where tonality may be dependent on the scene, it is not always desirable to produce linear tone curves as they may often appear "dull" due to reduced contrast. If you would like your profile to correct hue errors without affecting tonality (brightness or contrast), see the "Gamma Match" options below.
- Do Not Normalize: This option should be used when profiling a calibrated scanner that has a consistent (not automatic) exposure every time. If you have software that can produce a "raw" scan, for example, where exposure is identical in every scan, use this option. Since "Do Not Normalize" tries to exactly match the intended tone reproduction curve based on the shot of the test target, you wouldn't want to use this option when profiling a camera because you are unlikely to get the exact same exposure when you shoot different scenes.
- Gamma Match: Select one of the following Gamma Match options if you would like your profile to correct hue errors such as color shifts or saturation problems without affecting tonality. When profiling many scanners and cameras, it is often desirable to be able to correct problems such as reds shifted to orange, purples turning blue, undersaturation of yellow, and other color shift/saturation problems while letting the device and its user settings determine the tone curve. Most cameras/scanners use tone reproduction curves that include "enhancements" to make images more vibrant in order to obtain a "film like" quality. The "Normalize" and "Do Not Normalize" options try to correct the curves in order to make them linear/accurate and the result is often a dull or hazy appearance.
- Gamma Match (Auto): Being able to produce a profile that corrects hue shifts and saturation without affecting tone curves depends on the ability to discover the "base" tone curve; that is, what the tone curve should have looked like before the device altered it to make the photos have more pop and vibrance. For example, the camera being profiled may have started with a typical gamma 2.2 tone curve and then altered that curve to make the shadows a bit darker and the highlights a bit brighter. In order to reproduce tone curves as the camera intended (without altering them), Profile Prism must find/use the underlying gamma 2.2 curve as the tone reproduction curve in the profile. Some devices (and raw conversion software) use an underlying tone reproduction curve of gamma 2.2, others (typically raw conversion software) use gamma 1.8 or 1.0 when color management is turned off as it would be when you create a profile. Profile Prism should be able to detect the underlying/base tone reproduction curve when you select Gamma Match (Auto). In most cases, you will not need the manual options below, but if you know the underlying/base tone reproduction curve and Profile Prism is having trouble detecting it (because the curve was modified too much by the device to be able to detect it), use one of the manual Gamma Match options below.
- Gamma Match (2.2): Use Gamma Match (2.2) if you know that the base tone reproduction curve used by the device is gamma 2.2. If you are profiling in-camera JPEG's or images that have been saved to a common work space like Adobe RGB or sRGB, or even ProPhoto RGB, the base curve will be gamma 2.2. All but a few of the common working color spaces are gamma 2.2 so if your images have been saved with an embedded color space, it is likely gamma 2.2.
- Gamma Match (1.8): Use Gamma Match (1.8) if you know that the base tone reproduction curve used by the device is gamma 1.8. Some raw conversion tools (like Capture One when using "Embed Camera Profile") and a few rarely used working spaces like Apple RGB and Colormatch RGB use gamma 1.8.
- Gamma Match (1.0): Use Gamma Match (1.0) if you know that the base tone reproduction curve used by the device is gamma 1.0. Some raw conversion tools (like Bibble with color management turned off) and some raw scans use gamma 1.0. If the image of the target looks extremely dark when you open it in Profile Prism, chances are the base gamma was 1.0.
- White Balance: A common problem with developing profiles for digital cameras is white/gray shift due to inexact alignment of the camera's red, green, and blue tone curves. Even if white balance is 100% accurate for the target shot, the red, green and blue primaries may only align exactly in certain areas of the curve. In other areas, the curves may not line up exactly, causing a very slight color cast in what should be color-free gray areas. For example, take a look at the tone reproduction curves below:
On the lower part of the graph (in the shadows), the green channel is a bit higher than red/blue but in the brighter range, the red channel is higher than green/blue . In this example, the target shot would appear to have a reddish magenta cast on the darker gray patches along the bottom row of the reference target, and a slight cyan cast toward the highlights in that row. There are several things that can attribute to misalignment of curves such as color reflections on the target, lens chromatic distortions, slight errors in the camera's curves in firmware, etc. The problem with allowing profiles to generate individual tone curves is that the same camera used to capture the above target shot probably will not have the same misalignment every time. On the next shot, the red curve may be higher in the shadows and the blue curve higher in the highlights. Since the curve displayed above will always add some green to the shadows and red to the highlights, using this profile on a wide variety of images and lighting conditions can cause neutral tones to take on a color cast, although it can be used to completely neutralize grays/white under the exact same conditions as those that existed when the test shot was taken.
One way to prevent a profile from introducing unwanted color casts in neutral subjects is to use a single tone curve for all three color channels. By using the green (luminance) channel to generate the tone curve and ensuring that the red/blue channels follow this same curve, we can allow the device to dictate white balance by not changing what the device reports as neutral. In most cases, we want a single tone curve for digital cameras since they are always changing their tone curves under different lighting. In other cases like that of a scanner with a consistent light source, we might want to generate separate tone curves for all three RGB channels because this can correct for a device that might have a consistent color shift in neutral subjects.
Settings and their usage:
- Device Dictates WB: This is the default setting. This setting is appropriate for digital cameras and devices that have inconsistent tone curves for the three RGB channels. Use of this setting will prevent color casts from appearing in neutral subjects because all three channels will follow a single curve. A profile generated under this setting will not change grays/white: what is neutral in the input image will be considered neutral in the profiled image.
- Allow Profile to Fix WB: This setting should be used when the device you are profiling renders relatively stable tone curves (the RGB channels may not match, but they don't change from image to image). Scanners normally fit into this category since they have a static light source. You can also use this setting to correct images from a scanner or camera that typically renders neutral subjects with some amount of unwanted color cast. A profile generated under this setting will neutralize color based on the target shot.
- Brightness: Choose the "Normal" setting if you would like your profiled images to have an ideal tone curve (100% accurate brightness). Set this parameter to a positive value if you would like to increase brightness and shadow detail, and a negative value if you would like your profile to render images with darker shadows and an overall darker appearance. Note that the brightness here actually operates on a gamma curve, so a more accurate name for this parameter would be "gamma". The term "Brightness" is used simply so that it is a more recognizable and less confusing term.
- Contrast: Choose the "Normal" setting if you would like your profiled images to have an ideal tone curve (100% accurate contrast). Set this parameter to a positive value if you would like your profile to render images with higher contrast, and a negative value if you would like your profile to render images with less contrast. Although the "Normal" setting will ensure accurate contrast in the profile, some photographers may prefer higher contrast results to add more "pop" to images and increase separation between foreground and background. Most digital cameras and film processors deliver pictures with higher than normal/accurate contrast. Higher contrast tends to hide noise in the shadows and give a more "intense" or "sharper" feel to images, however, it also tends to destroy some amount of shadow detail. Since anything other than the "Normal" setting is considered an "effect" or "enhancement", the other values are available purely as a personal preference since many people prefer more contrast than provided by a totally accurate tone curve. The "Normal" setting also provides the greatest possible detail in the image, so if your normal workflow includes any type of post processing using a photo editor, it may be best to use the "Normal" contrast setting since this will ensure that the image has maximum detail.
- Saturation: Choose the "Normal" setting if you would like your profiled images to have ideal saturation. Set this parameter to a positive value if you would like your profile to render images with more saturation, and a negative value if you would like your profile to render images with less saturation.
- Bias Settings: The red, green, and blue bias values may be used to control unwanted color casts or to introduce different coloring effects. For example, if the profile produces a noticeable yellow cast, setting the red and green bias settings to negative values will reduce yellow in the profile. Bias changes can be made to the entire profile (if "bias includes grays" is checked) or only to non-neutral colors (if "bias includes grays" is not checked). Cameras can suffer from hue shifts when shooting under different light sources, while digital prints can be affected by metamerism (prints having a different appearance under different light), paper brighteners, and other issues. These are just a few examples where bias settings can come in handy by allowing manual "tweaks" when necessary. Note: always create profiles with bias settings at zero and only change the color bias in your profile if you deem it necessary.
- Reference Target: This reference file identifies the actual color for all color patches on your reference target and is how Profile Prism knows the actual colors on your target. Simply look at the name of the data file listed here and make sure that the same name/number is listed on your reference target in the lower right corner. Note that the target "version" is sometimes referred to as the target "charge". If the version/charge listed in the dropdown does not match the version/charge listed in the lower right of the reference target, select the proper file name that matches. Once you have verified this entry, it is not necessary to change it unless you start using a different paper target that has a different version number stamped in the lower right corner.
Step 2: Open and crop the image of the target
- Click "File", "Open Image" and browse to the folder that contains the image of the reference target. Select one of the images and it will appear in the image crop area in the upper right of the window.
- Next, locate the upper/left edge of the target in your image. To do this, use the horizontal/vertical scroll bars on the bottom or right of the image to scroll, or simply click on the image in the window and drag it left/right/up/down using the hand.
- Next, click the upper left crop corner button.
- Your mouse cursor will now change to an upper-left box corner when you move the cursor into the image area. Move this corner to the very edge of the target, placing it at the upper left edge of the black rectangle that surrounds the row/column labels as shown below. The corner marker is shown below as a black/white dashed line.
- Once positioned here, left click to place the corner mark. You will notice a red corner mark on the target image. If the corner mark is not exactly on the outside edge of the black corner as shown, simply repeat steps 3 and 4 until placed properly.
- Repeat steps 3 through 5, locating the other three corners of the target and placing their corner marks appropriately. Note that the top two corner markers should be placed at the edge of the outer/black rectangle surrounding the text row/column labels and the bottom two corner markers should be placed below the gray scale. Properly placed corner markers are shown below:
- Once all four corners of the target have been identified, the four "corner buttons" will only appear when you hover over them with the mouse. When all four corner buttons are "deactivated" and your target evaluation messages appear in the "Messages" area, you'll know that you have finished the cropping step. In addition, Profile Prism will overlay white "punchouts" on each color patch to verify alignment. The white punchouts should appear within each individual color patch on the target.
Note regarding the "Jiggle corners" option: If the "Jiggle corners" option above the image is checked, Profile Prism will "jiggle" all four corner markers to try to obtain the most accurate target alignment. Note that this option may move the red corner markers so that they no longer align perfectly with the edges of the target. This is normal since obtaining the best overall/average alignment of the white punchouts may require moving some/all corner markers slightly. To place the corners manually without Profile Prism moving them, simply uncheck the "Jiggle corners" box and place the four corner markers again.
- Simply move around the target and make sure that proper alignment exists by ensuring that each individual color patch contains a white punchout and that none of the white punchouts appear to overlap into neighboring color patches.
- Finally, review your messages. At this point, Profile Prism has examined the image of the target and has evaluated white balance, exposure, and lighting of the target. The details of this evaluation are displayed in the "Messages" box on the lower portion of the window. Here is a list of messages that are displayed at this point:
- Histogram: The histogram that displays in the lower left corner of the window after cropping a target can provide useful information about the capture of your color target. The histogram shows the distribution of pixel brightness values from left (black) to right (maximum brightness). The higher the curve, the more pixels contain that level of brightness. Ideally, brightness should be distributed from left to right on the graph with no large "clumps" of data on the left or right. A large spike on the left of the histogram indicates that some shadow detail on the target was clipped (not visible because it is completely black or zero in one or more RGB color channels). A large spike on the right side of the histogram indicates that some highlight detail on the target was clipped (not visible because it is completely white or maximum in one or more RGB color channels). Depending on the device being profiled, it may be impossible to achieve an "ideal" histogram where brightness is distributed throughout the entire 0-255 data range, but try to minimize "clumping" of data on the left/right as much as possible. Below are some examples of "good" and "bad" histograms.
The above shows a typical histogram of the IT8 reference target. The matte CRxxxxxx target produces a similar histogram but with more rounded edges on the left/right. The red endpoint markers on the left/right are very short, indicating that only a very small amount of data is at minimum/maximum brightness. As long as your histogram covers nearly the entire graph from left to right and does not contain any large red spikes at the endpoints, the histogram is acceptable.
The above histograms show data that is not uniformly distributed indicating improper exposure of the image. The graph on the left, although it certainly captures the entire range of brightness values in the image, is underexposed. With the histogram compressed into the lower portion of the graph, less data range is available for profiling and the resulting profile may not be as accurate as it could be if the brightness values were more evenly distributed. The image that generated the left histogram above will look dark and dull.
The histogram in the middle shows two problems. First, the red line on the right (although relatively short), indicates that some pixels in the image were at their maximum brightness (255). Also, notice how the data doesn't start until about 1/3 from the left. This indicates that there are no dark pixels in the image which compresses the capture range and makes black look gray. The image that generated this histogram will appear with a "haze" since contrast is lowered by the absence of true blacks and dark colors.
Finally, the histogram on the right shows a very large red (maximum) indicator on the right, indicating that many pixels were at their maximum brightness. This image is very overexposed and will not profile accurately. In addition, the histogram on the right also shows a red spike on the left indicating that some pixels were at zero (minimum) which can indicate loss of shadow detail. The image that generated the right histogram above will look super bright and oversaturated as well as being very contrasty. Unfortunately, images like this are what many scanners/cameras capture when set at their default values. Such images will have very vibrant color and a high level of "pop", but are far from realistic and will not profile well because much of the color target (in the shadows and highlights) cannot be captured due to data/pixels being "off the scale" by being too dark or too bright.
When possible, all of the above conditions should be avoided/minimized by tweaking the camera controls or raw conversion software controls if possible. The problem must be corrected up front (at capture time) and must not be corrected with photo editors after the fact since this cannot bring back clipped data that was lost due to an overexposed shot. Overexposure or clipping can normally be addressed by using exposure compensation on your camera.
- Exposure: Proper exposure is judged by the brightness of the white patch on the target (bottom right of the matte CRxxxxxx reference target). An ideal exposure will return RGB values of about 240,240,240 for that patch. Profile Prism will report an underexposure if the values drop below about 220 and an overexposure above 250. If the image is overexposed, some values could be clipped (beyond the recording range of the camera) causing problems with highlights. If the image is underexposed, the image could look washed out when the profile is used because only part of the camera's tone curve was captured. Note that since cameras try to meter the entire scene to middle gray, most cameras will underexpose the target shot, making it necessary to add exposure compensation to get values in the range needed. In addition, the firmware in some cameras actually modifies the shape of the tone curves based on the scene (what it "sees" in the frame). For this reason, we recommend shooting the target against a white background and using exposure compensation to get exposure to the proper level. See the next bullet for further details.
- White balance: Exact white balance indicates that there is no color cast to true grays/white on the target. Profile Prism will indicate whether white balance is inaccurate and to what extent. The percentage displayed is the amount that the RGB channels differ when evaluating a neutral color patch. Since neutral color patches should have equal RGB values, a 5% value here would indicate that there is a 5% color cast in neutral colors (a 5% difference between the minimum RGB value and the maximum RGB value for a neutral color patch). The higher this number, the more inaccurate the white balance. Inaccuracies in white balance will cause inaccuracies in the profile because there will be a color cast over the entire target. Note that as with exposure, cameras can be fooled by the many colors displayed in the target. If you cannot use a manual white balance or preset on a white/gray card, you may want to follow the recommendation of putting a large white sheet of (non glossy) paper behind the target. The white "border" provided when the target is attached to white paper can help cameras calculate the proper white balance and exposure.
- Lighting variance: Like white balance, the lower the number here, the better. This is the maximum variance in lighting measured on the gray luminance grid on the matte CRxxxxxx reference target. A value of 5% indicates that there is a 5% variance in lighting. That is, the darkest part of the target is 5% darker than the lightest. In a perfect shot, there would be no variance in lighting across the target. Lens vignetting and uneven lighting conditions make a theoretical best of 0% here nearly impossible to achieve. To minimize lighting variances, try to use a better light source, a light source that is farther from the target, and/or reduce the size of the target in the capture frame to reduce vignetting.
- Pixels at min/max brightness and clipping:
Pixels at min/max brightness: When Profile Prism examines an image of a target, it will determine the range of brightness values captured in the target, that is, the darkest and brightest pixel values found on the image of the target. Due to limitations of the capture device or simply due to the brightness range found on the target, these minimum/maximum values need not always be 0 and 255. An image of a target, for example, may range from 4-251, meaning that the darkest pixels found were a brightness of 4 and the brightest were 251. Profile Prism will report the number of color patches that contained at least some pixels that were at minimum or maximum brightness. Obviously, there must always be a darkest and brightest patch on the target, so at least two patches will always be reported as containing these darkest/brightest pixels. If a large number of color patches are marked as containing the darkest or brightest possible values however, it may indicate a problem. The message displayed will start with "Note:" if there are less than ten color patches containing min/max brightness and "WARNING:" if there are ten or more color patches that contain pixels at min/max brightness. A large number of patches that contain min/max brightness could indicate that the device that captured the image of the target simply "ran out of room" and bottomed/topped out. In a case where many color patches contained pixels of min/max brightness, since many pixels were found at min/max brightness, it is likely that at least some of them could not be properly recorded and the actual value could not be determined. Some things that can cause such a "truncated" range include: overexposure, oversaturation, inappropriate gamma setting, etc. If you get 10 or more patches at min/max brightness, try changing the exposure to see if it brings more patches into range.
Clipping: In contrast to color patches that contain some bottomed/topped out pixels, clipping can occur in more extreme cases. Clipping is a condition where all pixels in a measured color patch are at their min/max value or there was some anomaly in the measurement that indicates that the values in those patches cannot be depended upon to be accurate. There are three conditions that can cause clipping: (1) every pixel measured in the color patch was at min/max value, (2) no incidental variation (noise) was found in pixels: they were all measured at exactly the same value, or (3) no difference was found between two consecutive color patches. Any of these three conditions normally indicates that the value measured is probably not accurate. Once again, try altering exposure, saturation, or gamma settings to compensate. The number of clipped patches (marked with a black "X" marker) should be minimized since this indicates that the entire target could not be captured and may result in less accurate profiles.
"X" markers and what they mean:
Green "X": A green "X" indicates that some pixels were at the minimum value (shadow threshold). Since some pixels were at minimum value and some were not, this often indicates that the color patches in question are simply the darkest measured patches on the target. If only a few patches contain green "X" markers, no problem is indicated. If a large number of green "X" markers appear however, it could mean that the color is at or near the minimum brightness value detectible by the device. In the case of many green "X" markers, your capture device may simply not be detecting dark colors reliably and this could be an indication of underexposure, inappropriate gamma setting, or simply a limitation of the device being able to capture darker colors.
Red "X": A red "X" indicates that some pixels were at the maximum value (highlight threshold). Since some pixels were at maximum value and some were not, this often indicates that the color patches in question are simply the brightest measured patches on the target. If only a few patches contain red "X" markers, no problem is indicated. If a large number of red "X" markers appear however, it could mean that the color is at or near the maximum brightness value detectible by the device. In the case of many red "X" markers, your capture device may simply not be detecting bright colors reliably and this could be an indication of overexposure or oversaturation.
Black "X": A black "X" on a color patch indicates that the color patch was clipped. This condition is worse than pixels being at min/max value since it indicates a problem where inaccurate capture of the color patch is nearly guaranteed. See "Clipping" above.
Scrolling through clipped patches: To locate and scroll through the clipped patches on the target to view them, simply click on the "WARNING/Note: n patches are at minimum/maximum value..." message in the message box on the bottom of the window. Each time you click on the message, Profile Prism will move to the next clipped patch on the target.
Densitometer: To further assist with evaluating the colors in your captured target image, Profile Prism offers a densitometer that can be used to view the measurements for each color patch. After all four crop corners have been placed on a target image, you can point to any color patch on that target and right click to see Profile Prism's evaluation of that color patch. You will be able to view the original measured color of the patch. Using the densitometer, you can determine which color channel is clipping (if clipping occurred) and the exact color that was measured. Note that when examining color patches that are clipped with a black/green "X" marker, you will rarely see RGB values at their minimum/maximum value because you are viewing the RGB average over many pixels. Only some pixels were found to be clipped, so the average measured values will not normally show clipping.
NOTE: After you click "Create Profile" and the profile creation process has completed, the final messages displayed in the "Messages" box will be saved and associated with the image file that was used to create the profile. To recall the messages for a previous profiling session, simply click "File", "Recall Messages For" and select the image file used to create the profile. The messages will be recalled from the last time a profile was created based on that image.
Step 3: Generate the profile
Now it's time to sit back and let Profile Prism do its number crunching.
- Click the "Create Profile" button in the lower left corner.
- Note that Profile Prism will be performing billions of mathematical operations in the process of creating your profile, so it may take several minutes to generate a profile. You may follow progress by observing the progress bar at the bottom of the window.
- Note that once profile generation is complete, the location of your new profile is visible on the status bar at the bottom of the window.
- Make a final review of the messages displayed in the "Messages" box. Were any messages added? Profile Prism will add messages as the profile is being generated. These messages are:
- Spectral sample range: The spectral sample range is an indication of how "evenly" colors are spread across the spectrum by the device you are profiling. Profile Prism looks at different areas of the color spectrum when calculating a profile and needs sufficient measurements to ensure an accurate profile. Since this evaluation is reported as pass/fail, if a "Fail" is reported, it means that Profile Prism may not have enough information in certain areas of the spectrum to derive accurate color. "Pass" indicates that enough information is available for an accurate profile. A "Fail" on this evaluation means that there is at least one area of the spectrum where the device being profiled has a very narrow range: for example, very narrow range of greens, narrow range of orange hues, etc. Changing the "Profile For" parameter to "Smoother Gradients" will almost always result in a "Pass" on this evaluation if a previous attempt returned "Fail".
- WARNING - color patches detected out of correction range: Like an evaluation of "Fail" in the spectral sample range, this warning message is very rare and will normally not appear. It indicates that Profile Prism found color patches on the target that are so far off that it could not correct those colors completely. While some color patches may not be 100% accurate in the profiled image if this message appears, it doesn't mean that the error will be visible. In fact, the profiled colors will always be more accurate than the original. This message simply means that Profile Prism was only able to improve the colors but not able to make them 100% accurate. This message usually indicates that there are large errors in color rendering by the device in a small range of colors in the spectrum.
III. Utilizing Profiles Generated by Profile Prism
Note that the above instructions relate to generating an ICC profile for digital cameras. Since an ICC profile is a standardized method for describing how a camera records color information, using these profiles with images from your camera is a task left up to your imaging/editing software. Inexpensive software does exist which allows you to fully utilize ICC profiles for color management of images, batch conversion, etc. One such ICC aware application is Qimage, also produced by ddisoftware, Inc.. As stated elsewhere, consult the help or users manual of whatever ICC aware imaging/editing software you are using for assistance in making use of your ICC profiles once created. In Qimage, some relevant links that provide help and understanding of ICC profiles are:
Understanding ICC Profiles
Qimage Learn by Example: ICC Profiles Section
Note that whatever software you use, there is a standardized folder for which ICC profiles are normally stored. The following are "standard" locations for ICC profiles depending on the operating system used. It is recommended that you save your profiles in these locations since most software will look here for them:
Windows 95, 98, 98SE, ME: \windows\system\color
Windows NT/2000: \winNT\system32\spool\drivers\color
Windows XP: \windows\system32\spool\drivers\color