Abstract

The aim of the Computed Tomography (CT) process X-ray is to generate an image of an object from the energy attenuation. Thus, the linear problem that represents system acquisition, simulate the propagation of energy through the discrete object using over-determinate matrix. This system requires high computational resources, determining the time of reconstructions for iterative algorithms. However, the reconstructions estimate each pixel value as an individual variable, even if the CT image exposes similar features in a neighborhood of pixels. For this reason, this work introduces a multi-resolution strategy to reformulate the linear problem into a reduced set of unknowns that summarize the similar pixel values, enabling reconstructions 3.5x faster and up-to 16 dB of PSNR improvements compared with the traditional approach.