Depth variant image restoration in 3D fluorescence microscopy: two approaches under Gaussian and Poissonian noise conditions
Résumé
In this article, we are interested in restoring images from 3D fluorescence microscopy. In fact, these images are affected by a depth-variant blur due to light refraction phenomenon. We present and compare two different restoration strategies for that problem. The first one is based on multiple deconvolutions with depth-invariant blur functions and the second one consists in using a depth-variant blur function in the deconvolution process. Furthermore, we fit two deconvolution algorithms to this problem. First, we use the Richardson-Lucy method with total variation regularization to restore confocal microscopy images which are affected by a Poisson noise. Then, we focus on restoring wide field microscopy images which are corrupted by a Gaussian noise. Tests on simulated data show that the second restoration strategy is slightly more accurate than the first one for both noise conditions.