Phase imaging is especially important in biological imaging, where the samples are typically transparent (i.e. invisible) but yet introduce phase shifts that encode important information such as shape and density maps. In this talk, we develop linearized forward models for imaging transparent objects by using a set of images acquired at different axial positions. We first investigate the effect of the axial separation on the retrieved 2D phase maps. Based on this, we propose a practical phase reconstruction algorithm that combines different ranges of spatial frequencies—depending on the defocus distance of the measurements—in a regularized fashion. We illustrate and validate the method on real microscopy data of HeLa cells. We then extend our ideas to tomographic phase imaging and develop a 3D phase reconstruction algorithm that relies on the acquisition of a single image at each axial location. This enables a straightforward integration of fluorescence super-resolution optical fluctuation imaging within the same optical instrument. This imaging platform thus unifies the sensitivity of phase tomography with the specificity and high spatial resolution of fluorescence imaging, presenting a unique tool to study the complex spatial physiology of cells.