Focal spot shape estimation of computed X-ray radiography devices using multiple regions of interest on a single image of a tungsten sphere

The present contribution introduces an automated procedure for estimating the size and the shape of the effective focal spot of an X-ray source using a single 2D experimental image of a tungsten sphere. The proposed procedure is based on an iterative deconvolution scheme where the effective focal spot is estimated from an experimental image and a Digitally Reconstructed Radiography (DRR) model. The iterative scheme is only applied on small, automatically selected, regions of interest; allowing a dramatic reduction of the computation time. Moreover, another iterative scheme is used for accurately estimating the position of the sphere in the DRR model. This allows a simpler experimental setup for the acquisition of a 2D experimental image, as the position of the sphere need only to be known approximately. In order to assess the accuracy of the proposed procedure, sizes and shapes of effective focal spots were extracted from experimentally acquired cone-beam computed radiography (CR) images using X-ray sources with different focal spot sizes. These effective focal spots were subsequently used to deblur real-world radiographies using a fast deconvolution algorithm; showing that the proposed approach could significantly improve the resolution of CR images.