Local structural alignment is an effective method to detect the local similarities between two protein structures. Orengo & Taylor (1993) have already developed a local structural alignment method by using a double dynamic programming algorithm. In the method, the similarity between a pair of residues is evaluated as the similarity between the pair of the structural environments corresponding to the residues. However, it is difficult to evaluate the structural similarities between a pair of proteins related by circular permutation, because the structural environment of a residue drastically differs from that of the corresponding residue of the circularly permuted protein. In this manuscript, we propose a new method to construct a structural environment that is robust against circular permutation. We examined its efficiency in the detection of the local structural similarity by the reconstructed structural environments.