An important consideration for the successful use of fresh osteochondral allografts for the patellofemoral joint is selection of appropriately sized donor allografts. This is particularly challenging for matching grafts to the patella because of the substantial variability in articular contour that exists. At present, selection of donor patellae is based only on the width of the proximal tibia, with no established means of quantifying basic patellar size, facet configuration, or articular contour. We hypothesized that using a combination of standard radiographic measurements to generate an accurate profile of the recipients native patella would result in a more accurate donor graft. We radiographically measured seven dimensions of 30 fresh frozen cadaveric patellar specimens, and then, following dissection, accurate ex vivo measurements of patellar morphology were taken. Measurements obtained on the merchant view included total width, width of the medial and lateral facets, central ridge height (CRH) and location, as well as the patellar angle. The lateral view was obtained to measure the articular length. Ex vivo measurements were subsequently taken using a standard ruler, caliper, and goniometer following removal of all soft tissue attachments. To simulate a larger population of knees, we then used bootstrapping methods with simple random selection and replacement to generate a sample of measurements from 4,500 knees. In the bootstrapping process, measurements from one random specimen are added to the “population” at each iteration until the target sample size is reached. Data from this bootstrapped population were subsequently used to calculate statistically significant radiographic predictors of each ex vivo measurement, using a multivariate regression model with backward elimination. These predictors were then used to compare the calculated values to the actual ex vivo measurements for each dimension in our original 30 specimens' sample. Overall, these formulas were very successful in predicting ex vivo measurements. While further validation is warranted, we propose that these formulas can be utilized to select a more accurate donor.