We propose to directly process 3D + t image sequences with mathematical morphology operators using a new classification of the 3D+t structuring elements. Several methods (filtering, tracking, segmentation) dedicated to the analysis of 3D + t datasets of zebrafish embryogenesis are introduced and validated through a synthetic dataset. Then, we illustrate the application of these methods to the analysis of datasets of zebrafish early development acquired with various microscopy techniques. This processing paradigm produces spatio-temporal coherent results as it benefits from the intrinsic redundancy of the temporal dimension and minimizes the needs for human intervention in semi-automatic algorithms.