Crystallographic texture plays a crucial role in determining the mechanical behavior of polycrystalline materials, particularly in plasticity modeling. Pole Density Functions, constructed from (experimental) texture measurements, provide essential information about the preferred orientations of grains, which influence anisotropic mechanical responses. Accurate reconstruction of these functions is key to characterizing material properties and feeding homogenization models. To address this, we develop a customized Gaussian Process Regression (GPR) model for reconstructing Pole Density Functions, integrating spherical-periodic distance measures with conventional stationary kernels. Our approach adapts the GPR framework to capture localized texture features while ensuring physically meaningful reconstructions through a log-linear data transformation. This transformation enforces the non-negativity of both interpolated function values and stochastic intervals, improving uncertainty quantification. The model’s performance is systematically evaluated on synthetic texture datasets, assessing reconstruction accuracy, feature preservation, and uncertainty quantification compared to the conventional spherical harmonics approach. Beyond reconstruction, we extend this framework to study the homogenized behavior of Representative Volume Elements (RVEs) built using our developments. Additionally, we investigate how experimental measurement errors influence the choice of RVE size, aiming to quantify the impact of uncertainty on material characterization. These studies seek to establish guidelines for optimizing RVE selection while maintaining accuracy in homogenization analyses.