U-net

Phase aberration compensation via deep learning in digital holographic microscopy

The experimental results confirm that the trained CNN can accurately segment the sample from the background area of the hologram, and that this method is applicable and effective in off-axis DHM.