Colour correction is the process of converting RAW RGB pixel values of digital cameras to a standard colour space such as CIE XYZ. A range of regression methods including linear, polynomial and root-polynomial least-squares have been deployed. However, in recent years, various neural network (NN) models have also started to appear in the literature as an alternative to classical methods. In the first part of this paper, a leading neural network approach is compared and contrasted with regression methods. We find that, although the neural network model supports improved colour correction compared with simple least-squares regression, it performs less well than the more advanced root-polynomial regression. Moreover, the relative improvement afforded by NNs, compared to linear least-squares, is diminished when the regression methods are adapted to minimise a perceptual colour error. Problematically, unlike linear and root-polynomial regressions, the NN approach is tied to a fixed exposure (and when exposure changes, the afforded colour correction can be quite poor). We explore two solutions that make NNs more exposure-invariant. First, we use data augmentation to train the NN for a range of typical exposures and second, we propose a new NN architecture which, by construction, is exposure-invariant. Finally, we look into how the performance of these algorithms is influenced when models are trained and tested on different datasets. As expected, the performance of all methods drops when tested with completely different datasets. However, we noticed that the regression methods still outperform the NNs in terms of colour correction, even though the relative performance of the regression methods does change based on the train and test datasets.
RGB-guided spectral recovery algorithms include both spectral reconstruction (SR) methods that map image RGBs to spectra and pan-sharpening (PS) methods, where an RGB image is used to guide the upsampling of a low-resolution spectral image. In this paper, we exploit Matrix-R theory in developing a post-processing algorithm that, when applied to the outputs of any and all spectral recovery algorithms, almost always improves their spectral recovery accuracy (and never makes it worse). In Matrix-R theory, any spectrum can be decomposed into a component—called the fundamental metamer—in the space spanned by the spectral sensitivities and a second component—the metameric black—that is orthogonal to this subspace. In our post-processing algorithm, we substitute the correct fundamental metamer, which we calculate directly from the RGB image, for the estimated (and generally incorrect) fundamental metamer that is returned by a spectral recovery algorithm. Significantly, we prove that substituting the correct fundamental metamer always reduces the recovery error. Further, if the spectra in a target application are known to be well described by a linear model of low dimension, then our Matrix-R post-processing algorithm can also exploit this additional physical constraint. In experiments, we demonstrate that our Matrix-R post-processing improves the performance of a variety of spectral reconstruction and pan-sharpening algorithms.