Purpose: The image registration of optical scans to radiographic images is essential for performing computer-guided implant surgery. This study aimed to evaluate the effect of different image matching conditions on the accuracy of image registration for computer-guided implant surgery in completely edentulous jaws. Materials and Methods: The optical scan image of a completely edentulous study model was registered to the respective cone-beam computed tomography data using three different image matching conditions: small point (SP), large point (LP), and entire surface (ES). For the SP and LP groups, gutta-percha markers (1.0 and 3.0 mm in diameter) were attached to a base template, and a radiopaque impression material was relined on the intaglio surface of template in the ES group. Image registration was performed by 20 operators in the images obtained from each group at an interval of 2 weeks (n = 20 in each group), and the registration accuracy was assessed by calculating the aligned position of the edentulous arch image. One-way analysis of variance with Tukey post hoc tests was used to compare the results among the groups (α = 0.05). Results: The mean registration error was significantly larger in the SP group (0.52 ± 0.19 mm) than in the LP group (0.29 ± 0.08 mm) and ES group (0.27 ±0.06 mm) (F = 24.689, p < 0.001). No difference was found between the LP and ES groups. The image matching discrepancy was more homogeneously distributed on the arch in the ES group than in the other groups. Conclusion: The accuracy of image registration is affected by the size of the congruent area shown in the optical scan and radiographic images. The entire surface-based matching method is more accurate as compared to the small point-based matching method in the image registration for implant planning in full edentulous jaws.