Carbon fibre reinforced plastic (CFRP) exhibits complex optical behaviour due to its anisotropy and highly scattering surface. These optical characteristics pose significant challenges for the automated laser-based inspection systems used in CFRP manufacturing, as they lead to variations in light interaction with the material, affecting the accuracy and reliability of inspections. To investigate this complex optical behaviour, an inverse optical model based on the Multi-Gaussian method has been developed. Laser speckle patterns from the CFRP surface are decomposed into multiple Gaussian components to model the material's optical properties. A greedy optimisation algorithm is employed to estimate the optimal coefficients for the Gaussian sets, which are further refined by introducing negative amplitude Gaussian components. These enhancements improve the optimisation, resulting in a better correlation between the Multi-Gaussian model and actual laser speckle measurements.