In fluorescence microscopy, acquired images suffer from the effect of blur and signal-dependent photon noise as well as signal-independent read noise. Image deconvolution is a widely used post-processing technique to recover these spatially blurred and noisy images. Although beneficial, deconvolution algorithms are not modelled to handle noise, leading to unnecessary artefacts and a low signal-to-noise ratio in the restored images. In this paper, we propose to overcome this inherent limitation using a three-step procedure. We first apply forward variance stabilization on the acquired data. We then use an optimization problem that handles noise and blur reduction. Finally, an inverse variance stabilization recovers the signal of interest. We perform experiments using synthetically corrupted test data and real fluorescence microscopy data to validate the potential applicability of our technique.