GS Iterative Algorithm for Phase-Only Fourier Hologram Calculation

The Gerchberg-Saxton (GS) iterative algorithm is a numerical method commonly used for computing phase-only Fourier holograms. Originally developed by Gerchberg and Saxton for phase retrieval problems in electron microscopy, this algorithm iteratively transforms and applies constraints between spatial and frequency domains to gradually approximate the target light field distribution.

In phase-only hologram calculation, the core concept of the GS algorithm involves knowing the intensity distribution of the target image (the image we want to reconstruct) while the phase information remains unknown. The algorithm implements an iterative process where amplitude constraints are applied in the spatial domain and phase constraints in the Fourier domain, ultimately yielding a hologram containing only phase information. In code implementation, this typically involves maintaining phase values while replacing amplitudes during Fourier transforms.

A standard GS iteration cycle comprises four key steps: First, combine initial random phase distribution with target amplitude; then perform Fourier transformation to the frequency domain; next, preserve phase information but replace amplitudes with uniform values in the frequency domain; finally, apply inverse Fourier transform back to the spatial domain and re-impose target amplitude constraints. This cycle repeats until convergence criteria are met. Programmatically, this requires efficient FFT operations and careful handling of complex arrays.

A distinctive feature of the GS algorithm is its ability to completely reconstruct target light fields using only phase information, making it valuable for applications in holographic displays and optical information processing. However, developers should note that the method may converge to local optima, and its convergence speed heavily depends on initial phase selection. Implementing random phase initialization with multiple restarts can help mitigate these issues.

Compared to direct calculation methods, the GS iterative approach offers superior handling of complex light field distributions and can improve reconstruction quality through additional constraints like multi-plane reconstruction. Modern enhanced versions incorporate various optimization strategies such as adaptive step sizes and hybrid constraints to accelerate convergence and avoid stagnation problems.