Illumination and surface reflectance are the fundamental elements of any virtual scene, which have a major effect on its overall appearance. In reality, although the human eye perceives the incoming light in the form of color, illumination and reflectance have an underlying spectral representation. Considering spectral information of the environment is essential for accurately modeling the real world optical processes, which has a significant role in computer graphics for producing qualitative results in various applications, such as lighting reproduction and spectral rendering. Spectral measurements for complex illuminants and materials are often either not publicly available or exist but are not suited for a particular case scenario. Practically obtaining accurate spectral data is a challenging topic in computer graphics due to the common trade-off between the excessive costs of the specialized equipment and the measurement accuracy. Thus, in this thesis we propose novel methods for acquiring and modeling spectral illumination and reflectance that simultaneously aim at practicality and at the quality of the recovered data. We first introduce our novel practical RGB-to-spectral upsampling method for estimating spectral power distributions of illuminants from RGB photographs, including the legacy RGB environments, without the requirement of any additional information. We then show how the spectrally upsampled RGB illumination can be employed for improved lighting reproduction using an LED sphere. We also present a novel practical method for measuring spectral skin reflectance using a controllable LED sphere and an off-the-shelf dermatological device. We demonstrate how to employ skin measurements to drive the spectral subsurface scattering model with appropriate complexity in order to match the appearance of real skin in photographs. Finally, we implement the spatially varying skin subsurface scattering in a physically-based renderer and validate our proposed measurement technique by generating 3D spectral facial renderings that are, for the first time, comparable to the photographs. The proposed methods are versatile and can potentially be adapted for the general use in spectral rendering pipelines, including real-time implementation, as well as for lighting reproduction applications employing different illumination setups.