GPU implementation of full vectorial Point Spread Function (PSF) fitting
Super-resolution microscopy is a technique in optical microscopy that enables one to go beyond the diffraction limit of light. At the core of super-resolution microscopy lies the fitting of the Point Spread Function (PSF). This is a compute-intensive task, but it is also highly parallelizable, making it ideally suited for modern computing architectures.
Previously, a model for the PSF was implemented in MATLAB. However, while this implementation gave highly promising results, the approach faced limitations in processing speed and memory usage. This, in turn, limited the achievable resolutions.
This project aims to develop a full vectorial PSF model for both multi-core CPUs and CUDA-enabled GPUs. With the project, we aim to significantly enhance the efficiency and accuracy of super-resolution microscopy. The outcome of this project is a scalable and versatile super-resolution microscopy software that is open-source and can be adapted to various research needs.
Kernel Tuner greatly simplifies the development of highly-optimized and auto-tuned CUDA, OpenCL, and C code, supporting many advanced use-cases and optimization strategies that speed up the auto-tuning process.