Regenerative therapies have recently shown potential in restoring sight lost due to degenerative diseases. Their efficacy requires precise intra-retinal delivery, which can be achieved by robotic systems accompanied by high quality visualization of retinal layers. Intra-operative Optical Coherence Tomography (iOCT) captures cross-sectional retinal images in real-time but with image quality that is inadequate for intra-retinal therapy delivery. This paper proposes a two-stage super-resolution methodology that enhances the image quality of the low resolution (LR) iOCT images leveraging information from pre-operatively acquired high-resolution (HR) OCT (preOCT) images. First, we learn the degradation process from HR to LR domain through CycleGAN and use it to generate pseudo iOCT (LR) images from the HR preOCT ones. Then, we train a Pix2Pix model on the pairs of pseudo iOCT and preOCT to learn the super-resolution mapping. Quantitative analysis using both full-reference and no-reference image quality metrics demonstrates that our approach clearly outperforms the learning-based state-of-the art techniques with statistical significance. Achieving iOCT image quality comparable to preOCT quality can help this medical imaging modality be established in vitreoretinal surgery, without requiring expensive hardware-related system updates.