A Phase Recovery Technique Using the Genetic Algorithm for Aberration Correction in a Coherent Imaging System

Yu Zhang, Hongwen Zhang, Guoqin YuanSensors 2023 · DOI

TL;DR

This paper introduces a novel method to correct aberrations in coherent imaging systems by recovering the exit pupil's phase information and applying a conjugate filter in the frequency domain. They enhance a genetic algorithm to more accurately solve for these aberrations, leading to improved image sharpness and detail without altering the optical path.

What Problem Does This Paper Address?

Achieving high imaging quality in optical systems often comes at the cost of increased complexity and size. This is particularly true for coherent imaging systems, where even minor aberrations can significantly degrade image clarity and visibility. The challenge is to improve image quality and system miniaturization simultaneously. Traditional methods might involve complex optical designs or require knowledge of the system's Point Spread Function (PSF), which can be difficult to obtain. This paper aims to address this by developing a computational approach to correct aberrations directly in the frequency domain, thereby enhancing image quality without modifying the physical optical setup.

Methods

The core of the proposed method lies in phase recovery and frequency domain aberration correction.

  1. Phase Recovery: The system leverages the principle that the image spectrum is the product of the object spectrum and the system's transfer function. By analyzing intensity measurements at the focal plane and a defocused plane, they can iteratively solve for the phase aberrations at the exit pupil. This is achieved by fitting the aberrations using Zernike polynomials, which are standard for describing optical aberrations.
  2. Genetic Algorithm (GA) for Optimization: To accurately determine the Zernike coefficients (which represent the aberrations), the paper employs a genetic algorithm. Recognizing the limitations of standard GAs, particularly the "precocity" phenomenon (premature convergence to a suboptimal solution), they introduce several improvements:
  3. Dual Competition: In the selection stage, a more robust competition mechanism is used to ensure better individuals survive.
  4. Parallel Recombination: The crossover operator is modified to perform parallel recombination, preserving more genetic diversity.
  5. Disaster Model: A "disaster model" is introduced for the mutation stage. If the algorithm stagnates for several generations, the mutation probability is temporarily increased to "shake up" the population and prevent getting stuck in local optima.
  6. Aberration Correction: Once the exit pupil aberrations are accurately determined, a conjugate aberration filter is constructed. This filter is applied in the frequency domain to the captured image, effectively canceling out the phase distortions and restoring image quality.

Datasets

The paper does not introduce new datasets. Instead, it relies on simulated data and experimental setups using standard imaging targets:

  • Simulations: The simulations utilize a "Cameraman" image as a test object.
  • Experiments: The experimental verification uses a "Gauss image light intensity distribution," which appears to be a pattern of geometric shapes designed to test imaging performance.

Key Insights and Ideas

  • Computational Correction is Powerful: The research demonstrates that significant image quality improvements can be achieved computationally, without altering the physical optical system. This is crucial for miniaturization and adaptability.
  • Enhanced Genetic Algorithm for Robustness: The "precocity" phenomenon is a well-known challenge in GAs. The paper's proposed improvements (dual competition, parallel recombination, and the disaster model) show a tangible benefit in achieving more accurate and robust aberration recovery, especially when dealing with complex aberration profiles.
  • Frequency Domain Filtering for Aberration Correction: Applying a conjugate filter in the frequency domain is an elegant way to correct phase aberrations, directly addressing the core issue without needing to know the system's PSF beforehand.
  • Defocusing Distance Matters: The study highlights that the choice of defocusing distance for data acquisition is critical. Too small a distance can lead to aliasing and inaccurate phase recovery, while too large a distance also degrades accuracy. There's an optimal range, not a simple linear relationship.

Key Results

  • Improved Image Sharpness and Detail: Both simulations and experiments show a noticeable improvement in image sharpness and edge detail after aberration correction.
  • Quantitative Improvement:
    • In simulations, the Energy of Gradient (EOG) increased by 14.25%, indicating enhanced sharpness.
  • In experiments, the EOG increased by 2.23%, confirming practical improvement.
  • Accurate Zernike Coefficient Recovery: The improved genetic algorithm demonstrated superior accuracy in recovering Zernike coefficients, especially for larger coefficient values, compared to the classical GA. For example, in one simulation with larger coefficients, the improved GA achieved a better Appraise Value (lower is better, indicating better recovery) of 686.7341 compared to the classical GA's 713.2161.
  • Reduced Residual Aberration: The residual distribution after recovery showed that the error was only about 1% of the original aberration.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the precision of phase recovery is not a simple linear relationship with defocusing distance; finding the optimal distance requires considering multiple factors.
  • While the improved GA is more robust, the "precocity" phenomenon can still be "very serious" to some extent, even with the improvements.
  • Future Work:
    • The authors suggest further studying the application of this aberration correction method to large field-of-view imaging optical paths to expand its applicability.
  • Exploring other optimization algorithms or further refining the GA could potentially lead to even higher accuracy and efficiency.

Summary updated

Abstract

For traditional imaging systems, high imaging quality and system miniaturization are often contradictory. In order to meet the requirements of high imaging quality and system miniaturization, this paper proposes a method to correct the aberration of coherent imaging optical systems. The method is based on the idea of phase recovery and the imaging principle of a coherent imaging system to recover the aberrations at the exit pupil of the system. According to the recovered aberrations, conjugate filters are constructed to correct the image quality in the frequency domain. The imaging quality of the system is improved without changing the original optical path, and the simplicity of the system is guaranteed. To solve the pupil frequency domain aberration more accurately, this paper adopts the dual competition and parallel recombination strategy based on the genetic algorithm and introduces the disaster model. The improved genetic algorithm can effectively restrain the appearance of the “precocity” phenomenon. Finally, the paraxial imaging optical path is simulated and verified by experiments. The results show that, after aberration correction, the image sharpness is improved and the edge information is richer, which verifies the feasibility of the coherent imaging system image quality enhancement method proposed in this paper.
BibTeX
@inproceedings{zhang2023phase,
  author = {Yu Zhang and Hongwen Zhang and Guoqin Yuan},
  title = {{A Phase Recovery Technique Using the Genetic Algorithm for Aberration Correction in a Coherent Imaging System}},
  booktitle = {Sensors},
  year = {2023},
  doi = {10.3390/s23187679},
}
Figure