Abstract. This paper introduces a novel and fast method for person reidentification using features extracted from the appearance of individuals observed in non-overlapped fields of views in a network of surveillance cameras. The proposed method involves segmentation of silhouettes into meaningful regions, which is close to human visual categorization of colorful clothes, consequently obtaining better performance in various poses. The spatial features extracted from these areas that include color features contribute to the robustness of the method due to illumination changes. In addition, the use of the voting scheme reduces the computational complexity of the algorithm, thus yielding a fast algorithm.