A level-set method with the 3D multiplicative-additive model for CT volume data segmentation

Abstract

3D industrial computed tomography (ICT) technology plays a vital role in non-destructive testing and evaluation of complex components. However, CT images often suffer from artifacts and noise because the CT imaging system is non-ideal. To achieve high-precision segmentation of ICT volume data, a level set (LS) method with a 3D multiplicative-additive (MA) model was proposed. First, according to the cluster property of local intensity and the 3D-MA image model, we developed a local energy. The global energy was established by integrating the local energy. Next, the LS function was embedded into the global energy to form a variational LS energy functional. Finally, the proposed energy functional was minimized to achieve segmentation. The experimental results demonstrate that compared with the state-of-the-art models our model has superior accuracy. In particular, the measurement results of the forest ball also show that the proposed method excels in geometric measurement based on CT images.