X-ray phase-contrast imaging (PCI) is a nondestructive method to measure the coating thickness of tristructural isotropic (TRISO) fuel particles, with the advantages of generating sharp interfaces between two coating layers. However, the analysis of PCI images is hindered by noise and subject to operator-dependence. The aim of this study was to develop a measurement method that can exclude frustrating noise corruption and achieve computerized automation. The total variation (TV) algorithm was applied to reduce the unnecessary information, and the denoising results were optimized with the introduction of contrast-to-noise ratio (CNR). Contours of denoised images were automatically extracted using an adaptive Canny operator, employing the P-tile method. Thickness data, derived using the proposed method, were compared with those of ceramography, and the results were confirmed by performing the one-way analysis of variance (ANOVA) analysis. This measuring method allowed the denoising process to restore images with appropriate edge features and detect edges with more sharpness and continuity in minimal time. When evaluated against ceramography, there was a minor discrepancy (less than 10%) in all coating layers. The one-way ANOVA analysis exhibited the rejection of the significant differences in the thicknesses measured by the two aforementioned methods. The proposed method seems promising for practical application of such nondestructive measurements.