GPU-accelerated compressive holography.

Area of Science:

• Computational Imaging
• Signal Processing
• Computer Science

Background:

• Compressive holography enables efficient data acquisition but requires computationally intensive signal reconstruction.
• Traditional reconstruction methods often rely on iterative algorithms that are time-consuming on standard processors.
• Graphics Processing Units (GPUs) offer massive parallel processing capabilities suitable for accelerating such algorithms.

Purpose of the Study:

• To implement and evaluate a fast iterative shrinkage-thresholding algorithm for compressive holography signal reconstruction on a GPU.
• To leverage GPU parallel computing for significant speed improvements in compressive holography.
• To optimize the algorithm by exploiting matrix structure for enhanced computational efficiency.

Main Methods:

• Implementation of a fast iterative shrinkage-thresholding algorithm tailored for ℓ1 and total variation (TV) regularization on a GPU.
• Utilization of data-parallel computing principles inherent to GPU architecture.
• Exploitation of the measurement matrix structure to optimize matrix multiplication computations.

Main Results:

• The GPU-based implementation demonstrates substantial acceleration of the compressive holography signal reconstruction process.
• Achieved approximately 20 times faster reconstruction speeds compared to a CPU-based implementation.
• The parallel nature of the algorithm is effectively harnessed by the GPU for efficient computation.

Conclusions:

• GPU acceleration provides a significant performance boost for compressive holography signal reconstruction.
• The implemented algorithm offers a practical and efficient solution for time-critical holographic imaging applications.
• This approach paves the way for real-time or near-real-time compressive holography.