3D Zernike Descriptors

pyzernike is a Python library for computing 3D Zernike invariants as descriptors for 3D shape comparison.

Usage

Here's a basic example of how to use the library:

import numpy as np
from pyzernike import ZernikeDescriptor

# Create or load your 3D array here
arr = np.zeros((50,50,50), dtype = np.float32)
arr[15:25, 5:15, 35:45] = 1

# Fit the Zernike descriptor up to order 8, using data precision (float32 or float64).
descriptor = ZernikeDescriptor.fit(data = arr, order = 8)

# Get Zernike coefficients
coefficients = descriptor.get_coefficients()

# Reconstruct the array
arr_rec = descriptor.reconstruct(box_size = 50)

# Write the coefficients to a binary file
descriptor.save_invariants("coefficients.inv")

Installation

We recommend installation using one of the following methods

[!TIP] pyzernike uses OpenMP for parallelization. On ARM MacOS systems, brew install llvm libomp and adding the llvm binary to PATH is required prior to pip install.

Background

pyzernike provides Python bindings to C code written by Marcin Novotni, which was distributed under a GPL license and provided with the paper:

M. Novotni, R. Klein "Shape Retrieval using 3D Zernike Descriptors" Computer Aided Design 2004; 36(11):1047-1062

A copy of that code serving as the basis for this project was obtained from GitHub. pyzernike includes a range of modifications to the original code base to improve performance but faithfully implements the original derivations.