Diffractive optical network for reconstruction of holograms

A method of forming a physical embodiment of a diffractive optical network that reconstructs input holograms to reveal accurate optical images, optical signals, or optical data comprising:

• training with at least one computing device a diffractive optical network model formed by one or more diffractive layers to reconstruct at least one optical image, optical signal, or optical data input to the diffractive optical network as one or more holograms, the one or more diffractive layers having a plurality of optically transmissive and/or optically reflective physical features located in different locations therein or thereon, wherein the training comprises feeding an input plane of the diffractive optical network model with training holograms of optical images, optical signals, or optical data and computing an optical output of the diffractive optical network model through optical transmission and/or reflection resulting from the one or more diffractive layers and iteratively adjusting transmission and/or reflection properties for the one or more diffractive layers of the diffractive optical network model until optimized transmission/reflection coefficients are obtained to reconstruct the at least one optical image, optical signal, or optical data from the hologram(s) input to the diffractive optical network model, wherein the at least one optical image, optical signal, or optical data that is/are reconstructed is/are substantially free of twin-image noise or artifacts;
• and
• manufacturing or having manufactured a physical embodiment of the diffractive optical network model comprising one or more transmissive and/or reflective diffractive layers having physical features that match the optimized transmission/reflection coefficients obtained by the training of the diffractive optical network model.