Virtual tomography of a polymer gel's structure from a single SEM image

We present a computational framework for converting a 2D image obtained via electron scanning microscopy to a 3D object with reconstructed invisible interconnections. The approach consists of depth estimation using a perspective- and depth-of-field-corrected grayscale level, followed by the skeletonization of images corresponding to a set of equal depths. The obtained skeletons are further adjusted against their visible counterparts using the Gaussian convolution transform, and the resulting set of images is assembled into a stack of slices that satisfies the requirements of DICOM (Digital Imaging and Communications in Medicine), allowing efficient exploration of spatial structures via software developed for computed tomography. The approach is implemented as computer code and applied to porous structures comprising the polymer component of a recently created organogel exhibiting efficient chemomechanical behavior.