clij2-fft 0.29

A python wrapper around clij2 opencl FFT algorithms

clij2-fft (featuring non-circulant rltv deconvolution)

clij2-fft is a prototype implementation of a framework for OpenCl based FFT algorithms. The most used algorithm from this project is the OpenCL implementation of the non-circular Richardson Lucy deconvolution algorithm with total variation regularization, which can be called as follows

from clij2fft.richardson_lucy import richardson_lucy_nc
decon_clij2=richardson_lucy_nc(im,psf,100,0.0002)

or for large image that need to be split up into chunks with dask

from clij2fft.richardson_lucy_dask import richardson_lucy_dask
decon=richardson_lucy_dask(img, psf, 100, 0.0001)

If you need support for the library please post a question on the Image.sc Forum.

Long term we hope to integrate FFT based math more closely with the Clic project. The goal is to make it easy to write algorithms such as convolution, correlation, registration and deconvolution that consist of a series of FFTs combined with other math operations.