The utility of blue fluorescent protein (BFP) as a marker of gene expression has traditionally been limited by poor photophysical properties, including a low quantum yield and rapid photobleaching. We have used computational search algorithms to design libraries of BFP variants targeting positions neighboring the histidine-derived chromophore. The libraries were subsequently constructed and screened using flow cytometry. The fraction of functional clones in the designed libraries was considerably greater than in the fully-randomized libraries. Sorting the clones using FACS yielded several variants that were considerably brighter than the parental BFP, and the brightest variants contained previously unreported substitutions. Characterization of one of these variants revealed that it contained a considerably enhanced quantum yield (0.55 vs. 0.34) as well as improved photobleaching properties. The variants are well-suited for whole-cell cytometry, microscopy, or other fluorescent assays. These results suggest that computationally-focused protein libraries may overcome some of the limitations inherent in using either computational design or library screening approaches alone.