True orthorectification is a necessary process in the remote sensing realm. The discontinuity of object surfaces on ground would cause displacements and occlusions in images. Thus, the displacements and occlusions should be corrected and compensated, respectively, in the true orthorectificaiton. Considering the characteristics of ground surface and multi-layer roads, this paper presents a scheme for true orthorectification by using digital elevation model (DEM), road models (RM), and aerial images. The proposed scheme comprises three parts: (1) traditional orthorectification, (2) hidden detection, and (3) compensation for occlusions. First, orthoimages are generated with conventional orthorectification procedures. The second part is to detect hidden and double mapping areas. We design an H-buffer that determines if the corresponding image pixels belong to hidden areas. Thus, such pixels would be marked for further procedures. Finally, the marked areas are back-projected to images of another view-angle. The test data includes DEM, 3D road models, and multi-view digital aerial images. The test data covers a part of Taoyuan, in northern Taiwan. Experimental results indicate that the proposed scheme reaches reliable results.