Since the testing of convex mirrors often requires a compensator with a larger aperture than the mirror under test, this is undoubtedly a trade-off not worth pursuing. Moreover, for surfaces with higher levels of asphericity, it further escalates the challenges in testing. In this study, high-accuracy testing of large convex aspheric surfaces with rotary symmetry was achieved via a method that combines sub-aperture stitching and Computer-Generated Holograms (CGH) compensation testing. Applying the methodology outlined in this study, practical testing was conducted on a large-aperture convex aspheric mirror with a diameter of 538 mm. The outcomes demonstrated that the center testing data and the outer ring testing data were stitched by the stitching algorithm to obtain the surface error of the full-aperture. The results were compared to those of Luphoscan and compensation testing, with a residual Root Mean Square (RMS) less than λ/20 (λ = 632.8 nm). Finally, the error analysis of the entire testing process showed that the accuracy was better than λ/50. The results demonstrate that the method can achieve highly accurate test of large convex aspheric surfaces.