Abstract: Three-dimensional images are now generally formed by stacking a contiguous sequence of twodimensional slices．Typically，the spacing between slices is greater than that between pixels on a slice．A method to transform the slices into an isotropic volume data set is proposed．3D data are first classified into different matters， and then correspondence is established between pixels of the same matter in consecutive slices．The best matching pairs in these corresponding pixels are searched for using some matching rules，and interpolation between the two pixels of each pair is carried out．The method can effectively improve the interpolation quality，especially more clear scene at the boundary between different matters are obtained in comparison with traditional interpolation methods．