Functions (sigpy)¶
The core module contains functions and classes for signal processing.
SigPy provides simple interfaces to commonly used signal processing functions, including convolution, FFT, NUFFT, wavelet transform, and thresholdings. All functions, except wavelet transform, can run on both CPU and GPU.
These functions are wrapped into higher level classes (Linop and Prox) that can be used in conjuction with Alg to form an App.
Computing Backend Functions¶
Functions and classes for getting and setting computing devices.
sigpy.Device |
Device class. |
sigpy.get_device |
Get Device from input array. |
sigpy.get_array_module |
Gets an appropriate module from numpy or cupy. |
sigpy.cpu_device |
Device class. |
sigpy.to_device |
Move input to device. |
sigpy.copyto |
Copy from input to output. |
sigpy.Communicator |
Communicator for distributed computing using MPI. |
Block Reshape Functions¶
Block reshape functions.
sigpy.array_to_blocks |
Extract blocks from an array in a sliding window manner. |
sigpy.blocks_to_array |
Accumulate blocks into an array in a sliding window manner. |
Convolution Functions¶
Convolution functions with multi-dimension, and multi-channel support.
sigpy.convolve |
Convolution that supports multi-dimensional and multi-channel inputs. |
sigpy.convolve_data_adjoint |
Adjoint convolution operation with respect to data. |
sigpy.convolve_filter_adjoint |
Adjoint convolution operation with respect to filter. |
Fourier Functions¶
FFT and non-uniform FFT (NUFFT) functions.
sigpy.fft |
FFT function that supports centering. |
sigpy.ifft |
IFFT function that supports centering. |
sigpy.nufft |
Non-uniform Fast Fourier Transform. |
sigpy.nufft_adjoint |
Adjoint non-uniform Fast Fourier Transform. |
sigpy.estimate_shape |
Estimate array shape from coordinates. |
Interpolation Functions¶
Interpolation functions.
sigpy.interpolate |
Interpolation from array to points specified by coordinates. |
sigpy.gridding |
Gridding of points specified by coordinates to array. |
Pytorch Interop Functions¶
Functions for interoperability between sigpy and pytorch.
sigpy.to_pytorch |
Zero-copy conversion from numpy/cupy array to pytorch tensor. |
sigpy.from_pytorch |
Zero-copy conversion from pytorch tensor to numpy/cupy array. |
sigpy.to_pytorch_function |
Convert SigPy Linop to PyTorch Function. |
Simulation Functions¶
Functions for simulations.
sigpy.shepp_logan |
Generates a Shepp Logan phantom with a given shape and dtype. |
Thresholding Functions¶
Thresholding functions.
sigpy.soft_thresh |
Soft threshold. |
sigpy.hard_thresh |
Hard threshold. |
sigpy.l1_proj |
Projection onto L1 ball. |
sigpy.l2_proj |
Projection onto L2 ball. |
sigpy.linf_proj |
Projection onto L-infinity ball. |
sigpy.psd_proj |
Projection onto postiive semi-definite matrices. |
Utility Functions¶
Utility functions.
sigpy.resize |
Resize with zero-padding or cropping. |
sigpy.flip |
Flip input. |
sigpy.circshift |
Circular shift input. |
sigpy.downsample |
Downsample input. |
sigpy.upsample |
Upsample input. |
sigpy.dirac |
Create Dirac delta. |
sigpy.randn |
Create random Gaussian array. |
sigpy.triang |
Create multi-dimensional triangular window. |
sigpy.hanning |
Create multi-dimensional hanning window. |
sigpy.monte_carlo_sure |
Monte Carlo Stein Unbiased Risk Estimator (SURE). |
sigpy.axpy |
Compute y = a * x + y. |
sigpy.xpay |
Compute y = x + a * y. |