Fast Reconstruction of 3D Images Using Charlier Discrete Orthogonal Moments

摘要

We propose a new algorithm for accelerating the computation time of Charlier discrete orthogonal moments for three-dimensional (3D) images, based on two fundamental notions: The first is a new representation of 3D images called image cuboid representation (ICR) in which the 3D image is decomposed into a set of cuboids of the same gray level instead of voxels, enabling a considerable reduction in both the amount of treated voxels and the computation time of Charlier moments. The second is a matrix calculation of the Charlier moments instead of direct or recursive calculations. The significant reduction in the computation time for Charlier moments, in combination with the ICR method, motivated the development of this new method for 3D image reconstruction. Simulation results confirm the effectiveness of the proposed method in terms of the calculation time of 3D Charlier moments as well as the speed and quality of image reconstruction.