A Variable Neighborhood Algorithm Search Approach to Discrete Tomography from Noisy Projections with Applications to Industrial Gamma Tomography

E. A. O. Limaa, S. B. Melob, C. C. Dantasc

aEscola Politécnica da Universidade de Pernambuco, Recife-PE, Brazil

bCentro de Informática da Universidade Federal de Pernambuco, Recife-PE, Brazil

cDepartamento de Engenharia Nuclear da Universidade Federal de Pernambuco, Recife-PE,Brazil

Abstract

The industrial applications of tomographic image reconstruction methods still await for proper consolidation. The limiting factors include a wide variety of industrial technical and economic parameters, forcing reconstruction techniques to work with much less projecting views than the medical counterpart. In this context, we propose the application of Variable Neighborhood Search (VNS), a heuristic adapted to work on binary images, and at the same time keeping the control of the combinatorial explosion typical of discrete methods. Our results suggest that the application of this technique provides a useful compromise between computational complexity and robustness in the presence of random noise.

Keywords: Dinamic Tomography; Radon Transform; Non-Destructive Essays