Atlases provide a framework for information from diverse sources to be spatially mapped and integrated into a common reference space. In particular, brain atlases allow regional annotation of gene expression, cell morphology, connectivity and activity. In larval zebrafish, advances in genetics, imaging and computational methods have enabled the collection of large datasets providing such information on a whole-brain scale. However, datasets from different sources may not be aligned to the same spatial coordinate system, because technical considerations may necessitate use of different reference templates. Two recent brain atlases for larval zebrafish exemplify this problem. The Z-Brain atlas contains information on gene expression, neural activity and neuroanatomical segmentation acquired using immunohistochemical staining of fixed tissue. In contrast, the Zebrafish Brain Browser (ZBB) atlas was constructed from live scans of fluorescent reporter genes in transgenic larvae. Although different reference brains were used, the two atlases included several transgene patterns in common that provided potential 'bridges' for transforming each into the other's coordinate space. We therefore tested multiple bridging channels and registration algorithms. The symmetric diffeomorphic normalization (SyN) algorithm in ANTs improved the precision of live brain registration while better preserving cell morphology than the previously used B-spline elastic registration algorithm. SyN could also be calibrated to correct for tissue distortion introduced during fixation and permeabilization. Finally, multi-reference channel optimization provided a transformation matrix that enabled Z-Brain and ZBB to be co-aligned with acceptable precision and minimal perturbation of cell and tissue morphology. This study demonstrates the feasibility of integrating whole brain datasets, despite disparate acquisition protocols and reference templates, when sufficient information is present for bridging.