Fast and noise‐tolerant determination of the center of rotation in tomography

High‐quality tomographic reconstruction is not possible without the accurate localization of the center of rotation. Poor localization leads to artifacts in the data and can even cause reconstructions to fail. There are many approaches to solving this problem, some of which involve the collection of full sinograms, or even provisional tomographic reconstructions, in order to determine the center of rotation. Here, a simple method based on the expected symmetry of the Fourier transform of summed projections approximately 180° apart is presented; unlike cross‐correlation methods, it requires only a single Fourier transform to compute, and uses mainly low spatial frequency information which is less susceptible to noise. This approach is shown to be fast, and robust against poor signal‐to‐noise as well as to projection images acquired at angles that are not exactly 180° apart. This rapid method can be useful as a first step in the processing of tomographic data. A rapid phase symmetry method for finding the center of rotation in tomographic data using two images located 180° ± Δgθ apart is introduced. The method is more robust against photon noise than image correlation methods, and shows good tolerance to Δgθ ranging up to several degrees.