Over the past decade, novel live-imaging techniques have considerably changed our vision of cell biology, in particular in the field of neuroscience. Acquisitions of 3D image sequences over long periods of time, in particular, have enabled neurobiologists to follow complex processes such as the development of neuronal populations or degenerative events occurring in pathological contexts, improving our understanding of the mechanisms involved in brain development and function. In most cases, live samples are moving/growing during long-term imaging, therefore it is required to compensate for this global 3D motion before measuring the dynamics of the structure of interest. We present here a method to compute a coherent 3D motion over a whole temporal sequence of 3D volumes, which is able to capture subtle sub-voxelic displacements.