Parallel data processing for sparse data tomography sensors

The physical constraints encountered by Tomography in Industrial Process environments often restrict the access to the imaged subject, generating data that is limited both in the angular and in the radial sense. We overcome this problem by employing the Sinogram Recovery Algorithm (SRA) for limited views Tomography, based on sinusoidal Hough Transform. We demonstrate the parallelization potentials of this algorithm targeting the implementation of an embedded system capable of executing acquisition, reconstruction and visualization. We demonstrate the parallelized SRA in MATLAB by simultaneous processing of all acquired angular projections; the results generated by this implementation exhibit a satisfactory match with those obtained from the sequential version. Pilot stages of the algorithm also have been implemented in a Field-Programmable Gate Array (FPGA), providing results that show the adequacy of the method to perform real-time imaging in an embedded system.