Thinning algorithms for three‐dimensional gray images and their application to medical images with comparative evaluation of performance

In this paper, thinning algorithms considering the density of three-dimensional (3D) images are proposed and their characteristics are defined. In the 3D thinning methods reported so far, only two-level images which could not use the density information have been treated. Moreover, they are adversely influenced by noise in the shape of the objects. The proposed method can apply a thinning algorithm not only to the shape information but also to the density information. We apply this method to artificial objects and to real blood vessel area images extracted from 3D CT images; we then compare our results with the results of the conventional thinning methods applied to binary images and evaluate them qualitatively as well as quantitatively. It is shown that if we apply our thinning technique suitably with allowance for density information, then the influence of shape noise can be made smaller and natural thinning results can be obtained.