The objectives of this study were to evaluate ex vivo the effects of resin infiltration on the areal surface roughness of natural non-cavitated proximal subsurface lesions with or without previous deproteinization and to determine differences between E2 and D1 lesions or between premolars and molars. Forty premolars and 40 molars with proximal carious lesions and macroscopically intact surfaces (International Caries Detection and Assessment System (ICDAS) II; code 2) were radiologically assessed and randomly allocated to four groups (with 20 E2 and 20 D1 lesions, respectively). In each group, 10 lesions were deproteinized (NaOCl; 1%) before etching (HCl; 15%) and resin infiltration (Icon). Areal surface roughness (Sa) at the most demineralized lesion part (DIAGNOdent) was evaluated topometrically before and after deproteinization, after etching, and after infiltration using focus variation 3D scanning microscopy. Pretreatment with NaOCl (n = 40) had no significant effects on Sa (p = 0.208), but resulted in significantly differing Sa values between premolars and molars after etching (p = 0.011). Regarding the effects between etching and baseline, significantly differing Sa values (p = 0.0498) were found for premolars and molars (n = 40/40); Sa after resin infiltration (compared to etching) differed significantly between premolars and molars (p = 0.009). No treatment regimen lead to differences among the radiological grades (E2 vs. D1; p > 0.106). Resin infiltration showed only minor effects on Sa values of etched subsurface lesions (p < 0.170) and did neither equal nor improve baseline surface roughness (p > 0.401) of the different tooth types. Deproteinization should be recommended before etching and infiltration, even if surface roughness of infiltrated advanced (pre-)molar lesions will not be improved.