Multi-scale decomposition model-based segmentation algorithm for CT image with metal artifacts

Despite recent progress in X-ray computed tomography (CT), metal artifact remains to be a major problem in CT image processing and significantly limits numerous important applications. To overcome this problem, a CT image segmentation algorithm using an image multi-scale decomposition model based on the total variation/L¹ (TV/L¹) model is proposed in this paper. Different metal areas are extracted using the multi-scale decomposition properties of the TV/L¹ model, which allows the separation of metal areas from artifacts areas. Segmentation experiments are performed on real CT images of a printed circuit board with a distinct metal artifact. Results show that the proposed method is as effective as the interactive segmentation method, and it does not require interactions.