|
lsst.ip.isr gf5f1c85443+e8e946ae08
|
Public Member Functions | |
| __init__ (self, im, kernel, threads=1) | |
| __call__ (self, im, kernels) | |
Public Attributes | |
| fftImageObj = pyfftw.builders.rfftn(im, s=shape, threads=threads) | |
| fftKernelObj = pyfftw.builders.rfftn(kernel, s=shape, threads=threads) | |
| ifftObj | |
A class that performs image convolutions in Fourier space, using pyfftw. The constructor takes images as arguments and creates the FFTW plans. The convolutions are performed by the __call__ routine. This is faster than scipy.signal.fftconvolve, and it saves some transforms by allowing the same image to be convolved with several kernels. pyfftw does not accommodate float32 images, so everything should be double precision. Code adaped from : https://stackoverflow.com/questions/14786920/convolution-of-two-three-dimensional-arrays-with-padding-on-one-side-too-slow Code posted by Henry Gomersal
Definition at line 501 of file electrostaticBrighterFatter.py.
| lsst.ip.isr.electrostaticBrighterFatter.CustomFFTConvolution.__init__ | ( | self, | |
| im, | |||
| kernel, | |||
| threads = 1 ) |
Definition at line 515 of file electrostaticBrighterFatter.py.
| lsst.ip.isr.electrostaticBrighterFatter.CustomFFTConvolution.__call__ | ( | self, | |
| im, | |||
| kernels ) |
Perform the convolution and trim the result to the size of the input image. If kernels is a list, then the routine returns a list of corresponding convolutions.
Definition at line 532 of file electrostaticBrighterFatter.py.
| lsst.ip.isr.electrostaticBrighterFatter.CustomFFTConvolution.fftImageObj = pyfftw.builders.rfftn(im, s=shape, threads=threads) |
Definition at line 524 of file electrostaticBrighterFatter.py.
| lsst.ip.isr.electrostaticBrighterFatter.CustomFFTConvolution.fftKernelObj = pyfftw.builders.rfftn(kernel, s=shape, threads=threads) |
Definition at line 525 of file electrostaticBrighterFatter.py.
| lsst.ip.isr.electrostaticBrighterFatter.CustomFFTConvolution.ifftObj |
Definition at line 526 of file electrostaticBrighterFatter.py.