X-rays image analysis for defects detection and characterization in metallic samples

A way to assess metallic samples quality is to locate possible defects and to characterize them by parameters like depth, surface and volume. X-rays are well adapted to achieve this quantitative analysis for apparent as well as for internal defects. The important absorption and the granularity of some metals lead to very noisy radiographies. The authors propose a robust three-step inspection method: defects detection and localization, precise defects shape determination and defects parameters measurement. The approach, based on the combination of two specific radiographies, and on a relevant regularization, reduces noise effects and leads to a robust method. The regularization is based on the introduction of relevant geometric information by adapted Markovian modelizations. The method has been tested on a wide range of samples and compared to a destructive inspection. The paper is illustrated by the results obtained, at the different inspection steps, for a particular sample