Compensation enhancement by the patch-based inpainting in off-axis digital holographic microscopy

Automated surface profile measurement with digital holographic microscopy and large mask inpainting networks

“…This digitization process enables convenient computer programming for subsequent digital holographic reconstruction. In the reconstruction step, the virtual image I F H (x, y) can be obtained from the +1-order spectrum of the Fourier transform I H (x, y) through the angular spectrum reconstruction algorithm [11]:…”

“…From equation (3), to obtain the true phase of the sample, the constructed phase and the aberration phase of the whole captured (observed) area are required. In numerical aberration compensation methods, φ a (x, y) is often approximated by the phase distribution of the sample-free region that composes of the flat background region and its proper extension (area fillings) to the sample region by different algorithms, including patch-based method [11], two-stage GAN. This phase distribution can be fitted by using the Zernike polynomial [25], written as:…”

“…The experimental sample we used was a silicon wafer. Initially, the silicon wafer surface was etched using an inductively coupled plasma etching device, resulting in the formation of various patterns [11]. Figure 3 shows a schematic diagram of the experimental setup for a pre-magnified reflective off-axis digital holographic microscope based on Michelson interferometry.…”

“…In off-axis DHM recording systems, the real image can be separated from the conjugate image and the zero-order image to obtain a high-resolution reconstructed image, rendering the off-axis DHM increasingly being used. However, due to the off-axis configuration and the use of microscopic objectives, the image reconstruction inevitably introduces phase aberration that results in tilting and spherical bending, thereby reducing image quality and leading to inaccurate measurement of 3D information [11]. Addressing these challenges and enhancing reconstruction accuracy necessitates effective phase compensation methods.…”

“…Researchers have turned to image restoration algorithms to enhance phase aberration compensation by synthesizing sample-free holograms, which enable the determination of exact aberration parameters of the digital phase mask. Traditional restoration algorithms, such as the patch-based method, exhibit limitations in addressing stubby regions with multiple vacant distributions, and they are less effective for large or multiple vacant areas, leading to blurred results and lengthy computational efforts [11].…”