Representing Spectral Functions by a Composite Model of Smooth and Spiky Components for Efficient Full-Spectrum Photorealism

We propose a new model called the “composite model” to represent spectral functions. This model is built on the idea of decomposing all spectral functions into smooth and spiky components, each with its own distinct representation. A smooth spectrum can be expressed with coefficients of a set of given basis functions, and the discrete spikes in a spiky spectrum with their locations and heights. For the smooth part, we propose re-sampling functions that are reconstructed from the coefficients in a linear combination to improve efficiency. Spectral multiplication is thus greatly reduced in complexity. This new model shows remarkable advantages in representing spectral functions with aspect to accuracy, compactness, computational efficiency, portability, and flexibility, and it has a great application potential in color science, realistic image synthesis, and color image analysis. Here we apply it to rendering images involving real spiky illuminants as well as objects with light dispersion. The composite model is shown to surpass other models in these applications.