GPU-optimized matrix method for phase retrieval with the transport of intensity equation

Area of Science:

• Optics and Photonics
• Biomedical Imaging
• Computational Science

Background:

• Quantitative phase imaging (QPI) enables non-interferometric optical phase retrieval.
• Current TIE-based methods struggle to balance accuracy and computational speed.
• Addressing the accuracy-speed trade-off is crucial for practical QPI applications.

Purpose of the Study:

• To introduce a GPU-optimized iterative phase retrieval algorithm based on the transport of intensity equation (TIE).
• To enhance the accuracy and reduce the execution time of quantitative phase retrieval.
• To validate the proposed method for imaging biological specimens.

Main Methods:

• Developed the matrix solution to the TIE (MS-TIE) using a matrix formulation in the Fourier domain.
• Implemented iterative phase retrieval leveraging GPU parallelization for accelerated computation.
• Validated MS-TIE through simulations with diverse phase/intensity patterns and experimental imaging of biological samples.

Main Results:

• MS-TIE achieves improved accuracy and significantly reduced execution time compared to existing methods.
• The GPU optimization enables efficient parallelization of the matrix formulation.
• Demonstrated superior performance in balancing accuracy and speed for phase retrieval.

Conclusions:

• The proposed MS-TIE method offers a significant advancement in quantitative phase imaging.
• GPU acceleration of the matrix-based TIE solution enhances computational efficiency.
• MS-TIE effectively retrieves optical phase information with improved accuracy and speed, suitable for biological applications.