This repository details a computed tomography imaging spectrometer (CTIS) simulator, which is designed to simulate the generation of a center zeroth-order and surrounding first-order diffraction spots of a diffractive optical element (DOE).
The CTIS simulator is explained in the supplementary material of our paper The hybrid approach - Convolutional Neural Networks and Expectation Maximization Algorithm - for Tomographic Reconstruction of Hyperspectral Images [1].
This repository contains the following MATLAB scripts
- em.m: Expectation Maximization (EM) algorithm (see function for details, inputs and output)
-
generateH.m: Generate a system matrix
$\boldsymbol{H}$ for a computed tomography imaging spectrometer (see function for details, inputs and output) -
demo.m: Demonstration file showing generation of system matrix
$\boldsymbol{H}$ , simulation of CTIS image$g$ and reconstruction of hyperspectral cube$f$ using the EM algorithm.
and .mat files:
- cube_HSI_colorchecker.mat: A
$400\times200\times216$ hyperspectral cube captured by our pushbroom hyperspectral imaging system. - halogen25_mat:
$1\times 25$ column vector containing the spectrum of the halogen lamps used as illumination for the wavelength range 400 nm to 740 nm. - wavelength25_mat:
$1\times 25$ column vector containing the wavelength axis for 25 spectral bands (400 nm to 740 nm) - sensitivity25.mat:
$9 \times 25$ matrix containing the diffraction sensitivity (diffraction efficiency of the DOE, transmission of optical system and sensor response) for the 9 diffraction order spots.
A CTIS system is described by the linear imaging equation:
where
M.J. Ahlebæk, M.S. Peters, W.-C. Huang, M.T. Frandsen, R.L. Eriksen and B. Jørgensen, “The hybrid approach — convolutional neural networks and expectation maximisation algorithm — for tomographic reconstruction of hyperspectral images”, J. Spectral Imaging 12, a1 (2023), DOI: 10.1255/jsi.2023.a1