Speckle suppression method for infrared digital holograms based on sparse object representation and noise diversity

Digital holograms can be severely degraded by a mixture of speckle and incoherent additive noise. The problem is more severe in the case of IRDH, due to the large speckle grain. In order to suppress the speckles, here we present the MLDH-BM3D, a method specifically suited to filter DH images that combines Multiple DH captures, named Looks, to the BM3D, a numerical filter conceived in order to take advantage from a very sparse representation of the object features. We show that the joint action of ML and BM3D overcomes their respective limitations and achieves quasi noise free DH reconstructions.