Access to eye care can be expanded with high-throughput, easy-to-use, and portable diagnostic tools. Phase mask encoded imaging could improve these aspects of the fundus camera by enabling computational refocusing without any moving parts. This approach circumvents the need to adjust lenses to compensate for refractive errors. We developed a computational fundus camera by introducing a holographic diffuser at the plane conjugate to the ocular pupil, resulting in a laterally shift-invariant point spread function. We demonstrate computational refocusing of a model eye fundus over a large range of defocus errors. We also show computationally refocused, color, in vivo, human fundus images with a $\geq$35-degree field-of-view (FOV). This technology could eventually be combined with the wavefront-sensing capabilities of phase mask encoded imaging to create a compact ophthalmic imaging system that simultaneously captures a fundus image and performs aberrometry.