fle_2d

Installing in conda environment

conda create --name fle python=3.9 pip
conda activate fle
conda install -c conda-forge pip numpy scipy joblib matplotlib finufft
pip install fle-2d

Installing using pip

pip install numpy scipy joblib matplotlib finufft fle-2d

Testing install

git clone https://github.com/nmarshallf/fle_2d.git
# Or download folder and unzip
cd fle_2d/tests/
python3 test_fle_2d.py 

If you find the code useful, please cite the corresponding paper:

Nicholas F. Marshall, Oscar Mickelin, and Amit Singer. Fast expansion into harmonics on the disk: A steerable basis with fast radial convolutions. SIAM Journal on Scientific Computing, 45(5):A2431–A2457, 2023.

@article{marshall2023fast,
  author = {Marshall, Nicholas F. and Mickelin, Oscar and Singer, Amit},
  title = {Fast Expansion into Harmonics on the Disk: A Steerable Basis with Fast Radial Convolutions},
  journal = {SIAM Journal on Scientific Computing},
  volume = {45},
  number = {5},
  pages = {A2431-A2457},
  year = {2023},
  doi = {10.1137/22M1542775},
}

Acknowledgements

We thank Yunpeng Shi for contributing a vectorized version of the code for tensor inputs consisting of multiples images.

Usage

Given an image represented by a 2D array of size LxL that you want to expand into the disk harmonic basis, first create a basis object by calling

from fle_2d import FLEBasis2D
L = 128         #replace this by the side-length of your image
bandlimit = L   #maximum number of basis functions to use
eps = 1e-7      #desired accuracy
fle = FLEBasis2D(L, bandlimit, eps)

Here, eps is the accuracy desired in applying the basis expansion, corresponding to the epsilon in Theorem 4.1 in the paper. "Bandlimit" is a parameter that determines how many basis functions to use and corresponds to the variable lambda in equation (5.1) in the paper, scaled so that L is the maximum suggested.

All arguments to FLEBasis2D:

• L: size of image to be expanded

• bandlimit: bandlimit parameter (scaled so that L is max suggested)

• eps: requested relative precision

• expand_eps: requested approximate relative precision in the expand method (if not specified, pre-tuned values are used)

• expand_alpha: requested step-size in the expand method (if not specified, pre-tuned values are used)

• expand_rel_tol: requested relative tolerance in the expand method (if not specified, pre-tuned values are used)

• maxitr: maximum number of iterations for the expand method (if not specified, pre-tuned values are used)

• maxfun: maximum number of basis functions to use (if not specified, which is the default, the number implied by the choice of bandlimit is used)

• mode: choose either "real" or "complex" (default) output, using either real-valued or complex-valued basis functions

• precision: choose either "double" or "single"

To go from the image to the basis coefficients, you would then call either

coeff = fle.evaluate_t(image)

which applies the operator $\tilde{B}^*$ in Theorem 4.1 of the paper, or

coeff = fle.expand(image)

which solves a least squares problem instead of just applying equation $\tilde{B}^*$ once. The latter can be more accurate, but takes a bit longer since it applies evaluate_t maxitr times using conjugate-gradient (Richardson iteration is also implemented).

Once you have coefficients coeff in the basis, you can evaluate the corresponding function with expansion coefficients coeff on the LxL grid by running

image = fle.evaluate(coeff)

which corresponds to applying the operator $\tilde{B}$ in Theorem 4.1 in the paper.