An intelligent algorithm is proposed to extract the fascicular groups regional image from the nerve slice image. First, some pixels in the original image are selected arbitarily and their gray level comatrix with different neighbourhood scale are calculated in order to establish the texture feature matrix. Second, the rough k-means is employed to cluster all the arbitarily selected pixels. The arbitarily selected pixels which are located in fascicular groups area would be classified in the same cluster. Then, the convex hull of those pixels located in fascicular groups is calculated in order to extract the regional image. The experiments show that the new proposal not only can be quick and accurate to extract the fascicular groups regional image but is independent on modifying the parameter while processing other nerve slice images, which reflects the new algorithm has good adaptability.