Research shows that complex terrain can affect the spatial distribution of solar radiation and atmospheric physical processes. Based on the high-resolution topographic data, there are already several parameterization schemes available to couple the terrain effects on solar radiation with atmospheric models. However, to reduce the amount of calculation, some methods that can lead to errors are used in the sub-grid parameterization schemes for clear-sky direct solar radiation (SPS-CSDSR). In addition, the common finite difference slope algorithms and the assumption of consistent sub-grid atmospheric transparency can also result in errors. This renders existing SPS-CSDSRs unsuitable for complex terrain in middle and high latitudes and in turbid weather. In this study, these three problems have been effectively solved. The most accurate geometric algorithms for direct solar radiation so far, a high-precision and fast terrain occlusion algorithm and the triangulated sub-grid algorithm, are proposed. On Taiwan Island, the accuracy of the two methods is verified in the virtual vacuum atmosphere. Based on the fact that atmospheric transparency actually increases with altitude, a correction term based on sub-grid anomaly altitude is proposed for converting the sub-grid terrain effect factors into the atmospheric model. Overall improvements constitute a high-precision SPS-CSDSR in complex terrain. Eleven reduced calculation methods and common finite difference slope algorithms can no longer be used. In further study, atmospheric models need improvement in coupling the terrain effects on solar radiation to accurately describe vertical distributions. In this case, the high-precision scheme proposed in this study can play a key role.