A Variational Model For Poisson Gaussian Joint Denoising Deconvolution

In fluorescence microscopy, acquired images suffer from the effect of diffraction limited blur as well as signal dependent photon noise and signal independent Gaussian noise. Image deconvolution is a widely used post-processing technique which is used to recover these spatially blurred and noisy images. Although beneficial, the general ill-posed nature of the deconvolution problem often leads to unwanted artefacts in the restored image. In this paper, we consider the problem of restoring low SNR fluorescent microscopy samples from the perspective of using denoising coupled with deconvolution as a joint optimization problem incorporating a realitic PoissonGaussian noise model. Extensive comparison with different state-of-the-art deconvolution techniques, validate the superiority of the proposed approach.