This paper presents an automated approach for extracting vesse1s from computed tomography (CT) images that is based on也已mu1tisca1e ana1ysis of the vesse1 structure in也已image. According to也已sca1e間sponses to 1oca1line-like structure且，也已feature vesse1 image is built to reduce noise and enhance vesse1 strnctures. The mu1tisca1e linkage mode1 is used to build a pa間nt-child re1ationship between feature images with adjacent sca1es. Fuzzy distance measures are deve10ped to describe也已fnzzy simi1arity of vesse1 structures in mu1tisca1e image呂 立起fnzzysimi1arity is embedded into conventiona1 fnzzy clustering framework to build new mu1tisca1e vesse1 segmentation a1gorithm. The proposed mu1tisca1e similarity linking fnzzy C-means (MSLFCM) a1gorithm optimizes segmentation on all sca1es of也已featu間 images. The inter- and intra-sca1e cons個ints are combined to automatically extract vesse1s. Thesegmentation experiments are performed on syn也已tic and CT pu1monary vesse1 images. The experiment resu1ts demonstrate the satisfactory performance of the proposed approach.