Melhoria do desempenho de processamento de algoritmos de reconstrução de imagens em ensaios não destrutivos por ultrassom utilizando as ferramentas Cython e Numba

Abstract

This work aims to evaluate the use of Numba and Cython tools in optimizing the processing time of Python codes used for image reconstruction from nondestructive ultrasound inspections data. Only the image reconstruction algorithms that are based on the principle of delay-and-sum and that are widely used were evaluated: Synthetic Aperture Focusing Technique (SAFT), the Total Focusing Method (TFM) and the Coherent Plane Wave Compounding (CPWC). This paper proposes to use techniques to adapt codes, already developed in a framework, to improve runtime performance using the Numba and Cython tools. An evaluation of the implementations of these techniques is made by measuring and comparing execution times. An improvement of approximately 4 times for SAFT, 8 times for TFM and 388 times for CPWC was observed. All are the best results, and correspond to the use of Numba, but Cython also exhibited improvements in execution times with application of the techniques. This shows that the tools are effective in improving the runtime performance of image reconstruction algorithms.