As we work often with three-dimensional image data, potentially over time, classical image processing takes quite some time.
Hence, we will dive into image processing on graphics processing units (GPUs) using OpenCL, pyopencl, pyclesperanto and optionally using CUDA and cupy. This technology allows us to process images faster, GPU accelerated. Note that classical algorithms and GPU-accelerated implementations may differ in the very details.
In this chapter, we will also take a look at cupy, an NVidia CUDA based GPU-accelerated processing library and napari-cupy-image-processing, a scriptable napari plugin. These two can be installed using the following commands. This will however only work on computers that have a CUDA-compatible NVidia graphics card.
mamba install -c conda-forge cupy cudatoolkit=10.2
See also