The robustness of two tomography reconstructing techniques with heavily noisy dynamical experimental data from a high speed gamma-ray tomograph

The PSIRT (Particle Systems Iterative Reconstructive Technique) is, just like the ART method, an iterative tomographic reconstruction technique with the recommended use in the reconstruction of catalytic density distribution in the refining process of oil in the FCC-type riser. The PSIRT is based upon computer graphics' particle systems, where the reconstructing material is initially represented as composed of particles subject to a force field emanating from the beams, whose intensities are parameterized by the differences between the experimental readings of a given beam trajectory, and the values corresponding to the current amount of particles landed in this trajectory. A dynamical process is set as the beams fields of attracting forces dispute the particles. At the end, with the equilibrium established, the particles are replaced by the corresponding regions of pixels. The High Speed Gamma-ray Tomograph is a 5-source-fan-beam device with a 17-detector deck per source, capable of producing up to a thousand complete sinograms per second. Around 70.000 experimental sinograms from this tomograph were produced simulating the move of gas bubbles in different angular speeds immersed in oil within the vessel, through the use of a two-hole-polypropylene phantom. The sinogram frames were set with several different detector integration times. This article studies and compares the robustness of both ART and PSIRT methods in this heavily noisy scenario, where this noise comes not only from limitations in the dynamical sampling, but also from to the underlying apparatus that produces the counting in the tomograph. These experiments suggest that PSIRT is a more robust method than ART for noisy data. Visual inspection on the resulting images suggests that PSIRT is a more robust method than ART for noisy data, since it almost never presents globally scattered noise. (author)