NEW MICROFACET LIGHT REFLECTANCE MODEL BASED ON THE GGX DISTRIBUTION

In the article, it is proposed a new light reflectance model which is based on the GGX distribution. The developed model provides a highly realistic glare reproduction while simplifying the calculation of Fresnel and shadowingmasking coefficients. The main types of bidirectional reflectance distribution functions, which are used for glares formation, are analyzed. In particular, the microfacet bidirectional reflectance distribution functions and their components are considered. The key features of Gauss, Beckmann, Blinn, GGX functions usage for representing the microfacet distribution are described. The advantages of using the GGX distribution are listed. The standard expressions for Fresnel and shadowing-masking coefficient calculation are listed. It is shown that the existing expressions are characterized with a high computational complexity and not always correspond to the requirements of real-time rendering systems. The new piecewise polynomials for speeding up the Fresnel coefficient calculation are proposed. The performance gain from applying these expressions was obtained by calculating the total number of clock counts of the Intel Tremont processor. It is shown that proposed expressions provide a more accurate approximation of the reference function in comparison with the Lagarde expression. It is proposed the two approaches for speeding up the masking-shadowing coefficient calculation. The first approach involves choosing the optimal quadratic polynomial depending on the minimum upper value of the interval where the input angle parameter falls. The second approach involves the usage of the piecewise quadratic polynomials, which depend on the current parameter value. The plots of relative errors of obtained expression values from the original shadowing-masking function values are built. It is shown that the highly accurate approximation of the reference expression is provided. Based on the calculation of total clock counts of AMD Zen 4 processor, the productivity gain from applying the new shadowingmasking function formula is calculated. Additionally, the expression of simplified determination of function parameter is proposed. The proposed model is intended for use in real-time highly-realistic image formation systems.