Objective To enhance the opacity function for computed tomography (CT)-based prostate volume rendered (VR) images using reference magnetic resonance (MR) images. Materials and Methods After simulating on phantoms, ten patients with CT and MR datasets were identified/ registered. Optimal VR images were produced, and intra-modality and cross-modality errors [mean distance in voxel units (VU) from VR partial surface to target] were computed. On the CT VR images, a global transformation was computed to relate the opacity functions obtained using CT reference volumes with those obtained using MR reference volumes. Results The prostate was better visualized on MR than CT, as the intra-modality errors obtained using MR (2.78 ± 1.02 VU) were lower than on CT (3.68 ± 2.15 VU). Despite similar qualitative results as the intramodality CT images, the cross-modality errors obtained using the individual patient data (2.77 ± 0.99 VU) and using the global transformation (2.89 ± 0.99 VU) were each closer (P = 0.010 and P = 0.011, respectively) to the intra-modality MR errors than were the intra-modality CT errors. Conclusion By optimizing the opacity function using MR reference images, CT VR images of the prostate could closely represent the prostate defined on MR; furthermore, this process can be approximated by a simple transformation. Overall, the errors were reduced to a level (<3 VU) justifying further exploration in CT-based radiotherapy treatment planning.