A fast 3D euclidean distance transformation

A fast distance transform algorithm for 3D image is proposed. The 3D image is changed into slices of 2D images at first; then, two marked arrays for each 2D image are defined, according to which each pixel´s distance transformation in the 2D image is figured. Finally, the result of 2D distance transformation is applied to calculate each pixel´s distance transformation in the 3D image. The proposed algorithm can be easily implemented and parallel operated. The experimental results demonstrate the significant improvement in reducing time and space complexity when compared to boundary striping and Voronoi-based algorithms.