A method for total noise removal in digital holography based on enhanced grouping and sparsity enhancement filtering

In digital holography (DH), the coherent nature of the employed light sources severely degrades the holographic reconstructions due to a mixture of speckle and incoherent additive noise. These can affect both the visual quality in holographic imaging and display, and the accuracy of quantitative phase-contrast reconstructions. Typically, the noise problem is tackled by reducing the illumination coherence, thus the most intuitive way involves the recording of multiple uncorrelated holograms to be incoherently combined. This framework is known as Multi-Look DH (MLDH). However, single shot recordings are highly desirable in DH, and numerical methods are required to go beyond the improvement bound of ML techniques. Among the existing image processing methods, the 3D Block Matching filtering (BM3D) has shown the best performance. Here we present the MLDH-BM3D, a method specifically suitable to filter DH images that combines the two aforementioned strategies to overcome their respective limitations. We demonstrate the effectiveness of this framework in three different experimental situations, i.e. reconstructions of single wavelength holograms and color holograms in the visible region and the challenging case of the Infrared Radiation Digital Holography (IRDH) reconstructions, where a very severe noise degradation occurs.