Motion estimation, also known as optic flow, refers to the process of determining a 2D displacement field that aligns two images. Most methods that estimate motion or deformation fields in biological image sequences rely on sparse, distinct features (landmarks). Going a step forward, we are interested in methods to compute dense deformation fields (for all pixels). In this paper we compare two of such frameworks: the B-splines based free-form deformation (FFD) approach, which is well-known in medical image registration; and the combined local-global (CLG) approach, a popular optic flow method in computer vision. We test both methods on synthetic and real image sequences obtained by confocal light microscopy and by scanning electron microscopy, showing their performance in terms of accuracy and computational cost. As an alternative to traditional sparse techniques, the estimation of dense motion fields would allow tackling other related problems with sub-pixel precision, for example, the segmentation and classification of different biological structures according to their local motion, trajectory, growth and development.