We investigate methods for the recovery of reflectance spectra from the responses of trichromatic camera systems and the application of these methods to the problem of camera characterization. The recovery of reflectance from colorimetric data is an ill-posed problem, and a unique solution requires additional constraints. We introduce a novel method for reflectance recovery that finds the smoothest spectrum consistent with both the colorimetric data and a linear model of reflectance. Four multispectral methods were tested using data from a real trichromatic camera system. The new method gave the lowest maximum colorimetric error in terms of camera characterization with test data that were independent of the training data. However, the average colorimetric performances of the four multispectral methods were statistically indistinguishable from each other but were significantly worse than conventional methods for camera characterization such as polynomial transforms.

The proliferation of low‐cost colour imaging devices in the consumer market has led to a greater need to transfer images from one medium or device to another without loss of colour fidelity. A common solution is to characterise each device in terms of its CIE tristimulus values. In this paper two general techniques, artificial neural networks and polynomial transforms, are compared for their usefulness in characterising colour cameras. The neural and polynomial techniques are shown to give approximately similar performance once the parameters of the models are optimised. Since neural networks can be difficult and time‐consuming to train, it is concluded that polynomial transforms offer the better alternative for camera characterisation.

There is an increasing need to be able to measure colour properties of complex surfaces or images for which traditional spectrophotometers are not suitable. New multispectral imaging systems are being developed but it is not clearly understood how the parameters (such as the number of colour channels, the spectral properties of the channels, and the choice of illuminant) of such systems affect the performance. Furthermore, the effect of sensor and quantisation noise on the overall performance of the system also needs to be considered. This paper describes the development of a mathematical model of a multispectral imaging system that takes into account imaging parameters and noise. The results from the computational model show that increasing the number of colour channels alone in the imaging system does not necessarily allow better estimates of spectral reflectance. The choice of illumination can also, in the presence of noise, greatly affect performance.

A improved spectral reflectance reconstruction method is developed to transform camera RGB to spectral reflectance for skin images. Rather than using conventional direct or two-step processes, we transform camera RGB to skin reflectance directly using a principal component analysis (PCA) approach. The novelty in our direct method (RGB to spectra) is the use of a skin-specific colour characterisation chart with spectra closer to human skin spectra, and a new database of skin reflectances to derive the PCA bases. The experimental results using the facial images of 17 subjects demonstrate that our new direct method gives a significantly better performance than conventional, two-step methods and direct methods with traditional characterization charts. This new spectral reconstruction algorithm is sufficiently precise to reconstruct spectral properites relating to chromophores and its performance is within the acceptable range for maxillofacial soft tissue prostheses (error < 3 ΔE*_ab units).

In this paper we apply polynomial models to the problem of reflectance recovery for both three-channel and multispectral imaging systems. The results suggest that the technique is superior in terms of accuracy to a standard linear transform and its generalisation performance is equivalent provided that some regularisation is employed. The experiments with the multispectral system suggest that this advantage is reduced when the number of sensors are increased.