Implementation of the sequential Monte Carlo method in systems with massive parallelism for processing images in optical coherent tomography

Abstract

This paper discusses features of the use of systems with massive parallelism, using as an example graphic processors for the efficient implementation of the sequential Monte Carlo method (SMCM) in the case of data processing in optical coherent tomography. The dependence of the response rate of the SMCM implementations is studied in various software and hardware configurations. It is shown that it is expedient to use the OpenCL platform to implement the SMCM, including the case in which only a general-purpose processor is used. It is shown that computations are efficient on graphical processors, and this made it possible to process a three-dimensional tomographic image with a size of 1280×1022×376 readings in 30 sec, using the SMCM.