This document proposes a Bayesian Maximum a-Posteriori (MAP) method using sparse priors for 3D deconvolution of wide field fluorescence microscopy images of zebrafish embryos. The method uses a global Hyper-Laplacian prior to preserve sharp edges and a local smooth region mask to suppress ringing artifacts. Both synthetic and real zebrafish embryo microscopy data are used to evaluate the method, which demonstrates improved performance over state-of-the-art 3D deconvolution algorithms.