Reducing the variability in random-phase initialized Gerchberg-Saxton Algorithm

• Gerchberg-Saxton Algorithm with random initial guess produces different results.
• A new evaluation metric is presented.
• This new metric is able to measure the fidelity of the results with reduced variability
• Thus, the evaluation depends on the target rather than on the initial guess.

Gerchberg-Saxton Algorithm is a common tool for designing Computer Generated Holograms. There exist some standard functions for evaluating the quality of the final results. However, the use of randomized initial guess leads to different results, increasing the variability of the evaluation functions values. This fact is especially detrimental when the computing time is elevated. In this work, a new tool is presented, able to describe the fidelity of the results with a notably reduced variability after multiple attempts of the Gerchberg-Saxton Algorithm. This new tool results very helpful for topical fields such as 3D digital holography.