Camera based Colorimetry
A controlled illumination chamber can be used to provide appropriate setting for illumination and image capture of object colors. An example is shown in the Figure to the right. The unit shown contains two fluorescent lamps, along the sliding drawer, to provide the simulated illumination conditions. The object is placed on the surface of the sliding drawer. A calibrated digital camera is positioned on top of the chamber, the direction normal to the sliding drawer. The camera can be controlled by a computer to capture the image of the object automatically. Other designs are also becoming available.
The typical block diagram in Figure 1 shows the functional components of a digital still camera. Fundamental optics of digital still cameras are equivalent to those of color-film-based cameras. The output analog signals of the image sensor are processed by an analog preprocessor, where sample-and-hold, color separation, automatic gain control, tone adjustment, and other signal processing are carried out. These signals are then converted to digital signals via an A/D converter. The digital signals are processed by the digital signal processor and/or a microprocessor. The process includes tone adjustment, white balance and image compression. Image signals used for automatic exposure control, auto focus and automatic white balance are also generated here.
Many recent camera models feature RAW color coding. A camera sensor can only deliver one piece of information per pixel. For color imaging, at least 3 types of information (R, G and B signals) per pixel are needed. The sensors are usually arranged in a specific order, one of the most well known being the Bayer pattern. A Bayer pattern showing GBRG pattern is shown below.
Calibration of Camera
The calibration of the digital camera sets up the data mapping between the device space and the color space. There are different ways to compute the data mapping between the device space and the color space, such as regression, PCA, look-up-table (LUT), etc. A standard test chart is normally used. All the color patches in the standard test chart are usually previously measured with a spectrophotometer. Figure 3 shows the snapshot of a common test chart, the GretagMacbeth Colorchecker.
In most cases, a camera based system uses only three channels, i.e., R, G and B. More channels can provide additional information or constraints in calculations when higher color accuracy is desired. A multi-spectral system based on a monochrome camera plus 7 filters, as shown in Figure 4, may be used to closely approximate the reflected spectral curves of objects.
A well-defined and constant illumination would help minimize variability during image capture and analysis. The CIE has recommended 4 types of illumination and viewing geometries for measurement of the object color: (a) normal/diffuse (0/d), (b) diffuse/normal (d/0), (c) 45º/normal (45/0) and (d) normal/45º (0/45), as shown in Figure 5.