Study on registration and fusion methods for neutron/X-ray three-dimensional tomography images

Compared to two-dimensional neutron/X-ray computed tomography (CT) image fusion, three-dimensional (3D) fusion enables more comprehensive and accurate data analysis. Therefore, fusion method of neutron/X-ray 3D CT images was studied. In order to enrich information and enhance clarity of the fusion image, a NSCT decomposition-based image fusion method was firstly proposed, the neutron and X-ray images were decomposed into different scale spaces, and image fusion in different scale spaces was performed according to the fusion rules of taking average value, taking the maximum absolute value and the fusion rule of combining the dual-channel unit-linking PCNN (Pulse Coupled Neural Network) model, respectively. Then, using structural consistency between neutron and X-ray tomographic images of a tested sample, a structural similarity (SSIM) index-based inter-layer registration method for dual-source (neutron/X-ray) tomographic images was proposed. A 3D tomographic fusion method was thereby developed and volumetric image fusion was achieved. ‌In the 3D fusion images, the structural representations of plastic materials are predominantly derived from neutron tomographic image, exhibiting higher gray values, while metallic component information is primarily sourced from X-ray tomographic image, showing lower gray values. The fused images are synthesized by integrating the complementary information from both neutron and X-ray modalities, leading to enriched structural details that critically enhance material identification capabilities through cross-modal data fusion, which can be applied to the identification and detection of materials in multiple fields such as aerospace, electric energy, multiphase flow and composite materials in the future.