Although additive manufacturing enables the production of metallic components with complex geometries, accurately characterizing their surfaces remains challenging, especially in regions beyond the reach of conventional techniques. To address this challenge, this study investigates scanning acoustic microscopy (SAM) as an automated, non-destructive method for measuring the surface roughness of stainless steel 316L cubes fabricated by selective laser melting. A dedicated image-processing pipeline was developed to detect the water–specimen interface in SAM B-scan images and convert it into a calibrated surface profile. ISO-based spatial filtering was then applied to separate roughness from measurement noise, waviness, and surface form. The resulting profiles were used to calculate R_a, R_q, R_p, R_v, and R_t. R_a was used as the common parameter for quantitative validation of SAM against reference standards. For nominal values of 12.5, 25, and 50 μm, SAM measured 12.92, 28.53, and 51.60 μm, corresponding to absolute deviations of 0.42, 3.53, and 1.60 μm, respectively. The repeatability assessment showed coefficients of variation below 1%, indicating precise measurements. Following this assessment, the proposed pipeline was applied to the fabricated cubes and compared with digital light and laser confocal microscopes. Overall, the findings show that SAM can distinguish different roughness levels and provide reliable quantitative measurements.

Although additive manufacturing enables the production of metallic components with complex geometries, accurately characterizing their surfaces remains challenging, especially in regions beyond the reach of conventional techniques. To address this challenge, this study investigates scanning acoustic microscopy (SAM) as an automated, non-destructive method for measuring the surface roughness of stainless steel 316L cubes fabricated by selective laser melting. A dedicated image-processing pipeline was developed to detect the water–specimen interface in SAM B-scan images using robust surface echo tracking and convert it into a calibrated surface profile. ISO-based spatial filtering was then applied to separate roughness from measurement noise, waviness, and surface form, enabling the calculation of different surface roughness parameters. To evaluate the proposed method, R_a was selected for validation using seven reference standards with nominal values ranging from $0.8$ to $50\mathsf{\mu }\mathrm{m}$. The influence of Gaussian filtering was also investigated by comparing two formulations. SAM successfully distinguished all seven roughness levels, with absolute deviations ranging from $0.327$ to $1.772\mathsf{\mu }\mathrm{m}$ for the cutoff-matched formulation and from $0.199$ to $4.075\mathsf{\mu }\mathrm{m}$ for the simplified formulation. Furthermore, repeatability measurements demonstrated mean coefficients of variation below $2\%$ across all investigated specimens. Comparison between the proposed SAM pipeline, digital and laser confocal microscopes showed mean absolute deviations of $0.75$, $1.30$, and $2.85\mathsf{\mu }\mathrm{m}$, respectively, from shared reference values. Building on this comparison, the optical instruments were used to assess SAM measurements on fabricated cubes without known reference values. Overall, the findings support SAM as a complementary method for quantitative surface roughness evaluation.