Abstract
This research details a simulation pipeline methodology for applying a physics-based foundation simulation pipeline that leverages both full-spectrum Kubelka–Munk theory and the four-flux radiative transfer model. Unlike traditional alpha-blending, image-processing techniques, or contemporary deep learning methods such as GAN-based makeup transfer and neural style transfer, this approach simulates the actual physical light transport within a translucent cosmetic layer applied to human skin. While generative AI models often provide visually plausible results, they frequently lack physical consistency, fail to satisfy optical energy conservation (light transport) constraints, and cannot accurately account for complex 3D lighting interactions. By operating in a physically linear color space and reconstructing full-wavelength spectral reflectances from RGB data, this system ensures rigorous color accuracy and robustness under varying environmental lighting. This work focuses on the cohesive system-level synthesis of these physics-based components and provides a comparative empirical evaluation under professional DSLR and consumer mobile imaging environments. This provides a practical end-to-end pipeline that achieves physically consistent virtual try-on while remaining grounded in optical physics.

This publication has 13 references indexed in Scilit: