Segmentation of noisy CT volume data using improved 3D Chan-Vese model

The segmentation of the image directly using the CT volume data is attaching more and more attention. However, we can hardly get satisfactory segmentation results when the image is corrupted by noise. In this paper we propose an improved 3D Chan-Vese model (CV model) for the segmentation of noisy CT volume data. When working with level sets and Dirac delta functions in CV model, a standard procedure is to reinitialize level set function φ to the Signed Distance Function (SDF). At each iteration the SDF can be used perfectly to determine the type of filters with emphasis on removing noise or preserving the details. Thus the proposed model can suppress noise and preserve the contour at the same time. Experiments demonstrate that the proposed algorithm can effectively improve the segmentation of noisy CT volume data.