The Capture Process

This section highlights some of the pitfalls encountered during the capture process, both during profiling and reproduction.

Overview

A common method used to assess color in an image is to photograph a reference reflectance target (i.e., a color chart) that includes a set of color patches with specified characteristics and reference color values. The target is usually photographed at the beginning and at the end of the session, but it can also be included within each image. The intention is to calibrate the colors in the image so that they explicitly represent the original object being photographed.

In Capture One, the accuracy of the colors in the image is assessed by processing and comparing the measured color values using the readouts feature with the reference values for those patches.

Before measurement begins, it is necessary to adopt a color profile representing the capture set-up, or otherwise change the processing parameters, to ensure the processed image of the target is near to the stated reference. In many cases, the reference is specified in the CIE 1976 (L*, a*, b*) color space, often referred to as CIELAB, or more simply LAB, or just “Lab”.

For critical comparison, the capture process typically requires an accurate color profile to be adopted, either one created in-situ or a bespoke profile that includes the specific camera and illuminant (i.e., light source) for the set-up. Besides taking into into account the color temperature, measurement of the illuminant or light source must also include the geometry of the set-up (usually D50 at 45-degree angle of illumination), and the average human color vision (an observer model, usually the CIE 1931 2-degree Standard Observer). This device profile, and the one used during output for further analysis in third-party software, typically adopts an RGB color space.

Before looking at the challenges encountered when converting RGB to Lab, at least when validation or further analysis is required in third-party apps, lets look at some of the practical implications of adopting Lab as a reference color space.

A common way to assess color is to capture a reference reflectance target, which includes a set of patches with specified characteristics.

Lab and CIE XYZ color spaces

The Lab color space is itself derived from an earlier reference space, CIE XYZ. In 1931, CIE established a model based on an averaged observers’ visual sensitivity to different wavelengths of light under a specific light source and angle of illumination. From that model, the CIE introduced XYZ tristimulus values and when plotted in 3D form, three coordinates XYZ. In its 2-D form, color is plotted in an elongated n-shaped chart, known as the CIE 1931 XY chromaticity diagram.

Lab adopts a 3-D model that uses values that are easier to interpret, with L or Lightness co-ordinate and two a and b color components. The model also more closely matches human color vision, in respect to the perceived differences in color and the distances between them, especially when plotted in a 2-D form using circular color wheel. However, a Lab coordinate is computed from an XYZ coordinate by “normalizing” to a white-point. This means that under a certain light source, a color that is perceived to have the same color as white is neutral, and will have coordinates, or values, a*=b*=0. In Lab, that light source is D50.

Standard observer

The 2° Standard Observer models the center of normal human vision, which is the area most critical to color perception. An alternative is the CIE 1964 10° Standard Observer, that models a wider, and less color sensitive, area of human color vision. In practice, the choice of observer model only affects the measurement of the reflectance target when verifying color fastness using a spectrophotometer. As the Lab coordinate system does not impose any particular observer model, the choice has few consequences for the workflow, or other areas of color management.

Light source

Although a Lab coordinate is always specified with respect to a white point, it is fortuitous, or it is as a result perhaps, that flat art reproduction is generally standardized on the same D50 light source as specified in an ICC Profile Connecting Space (PCS). If a different light source is used or, more commonly, when using an ICC RGB profile with a different native white-point (or “media white”), you must convert between white-points. This process is called chromatic adaptation. Selecting profiles standardized on D50 such as ProPhoto RGB or ECI-RGB avoids this and, therefore, the need to support or interpret the required transform. However, there are some other practical limitations to overcome during conversion of ICC profiles that use a color-space with a different white-point, such as sRGB or AdobeRGB (1998).

Metamerism

Metamerism is perhaps the most challenging issue affecting the capture process. Objects that are perceived to have the same color under an observer model, are known as metamers. However, theses objects may not have the same perceived color when there’s a change in conditions, or in this instance light source and the angle of illumination. This is known as metamerimc failure. The consequence of this is that, to get the desired result, the conditions under which the capture of the color is acquired during both profiling and reproduction, must match very closely.

There are several reasons a set of metamers may have different colors when conditions change:

Observer metamerism
Since the observer model is specified, this is not usually an issue for reproduction. 

Illuminant metamerism
Color changes when the illuminant changes. In practice one should strive for light sources without any significant spectral peaks for high quality reproduction. Budget fluorescent tubes, energy-saving bulbs and LED lights can have spectral peaks that distort colors, even if they have the same color temperature (degrees K) as an expensive light source.

Instrument metamerism

This is caused by a mismatch between the observer model and the instrument. This needs to be calibrated, which is achieved using the camera profile in reproduction. Another example is monitor display calibration, using a simple tri-stimulus colorimeter. In this case, it might be necessary to stipulate a “display technology” manually (e.g., CCFL or LED) in the monitior calibration and profiling software. The mismatch between the instrument and observer can be so poor that the instrument fails the calibration process.

Geometric metamerism

This occurs when the geometry of illumination or viewing is changed.

By far the most challenging issue affecting the capture process is geometric metamerism. Ideally, a reflection target should consist of patches with a perfect reflecting diffuser. In this case, the appearance of a patch is unaffected by both the angle of the light and the viewing angle. In many cases it is unlikely that the same can be said for the objects or materials to be photographed. In practice, it is likely that there will be a significant specular component.

The specular component is highly sensitive to the angle of light and view. For colored materials, the hue of the specular component is usually closer to the hue of the light source than the hue of the diffuse reflection, with the result that patches becomes brighter (especially dark patches) and colored patches becomes less saturated (especially highly saturated patches). However, the hue of a patch can often be assessed accurately.

Note that, calibrating the processing to a set of test patches is only helpful to obtain accurate colors for materials that reacts similarly to light. For example, calibrating to a test target is great for calibrating test targets, but unlikely to be helpful to obtain accurate colors for glossy and translucent materials (e.g. paints, metals and porcelain), non-isotropic materials (e.g. textiles, papyrus and parchment) or textured surfaces (e.g. art with visible strokes and engravings).