As the nondestructive testing (NDT) images have the characteristic of unbalanced grayscale distribution, fuzzy cmeans clustering algorithm cannot commonly used effectively separate the objects from background in NDT images. To solve this problem, an improved suppressed fuzzy cmeans (ISFCM) clustering algorithm is proposed to segment the NDT images. Firstly, the total membership degree of each cluster is incorporated into the objective function of the suppressed fuzzy cmeans (SFCM) clustering algorithm, which can equalize the contribution of object pixels and background pixels on the clustering results. Secondly, the iteration forms of the new membership degree and cluster center are deduced on the basis of newly built objective function. Thirdly, the convergence of the proposed ISFCM clustering algorithm is analyzed and the implementation steps are given. Lastly, the proposed ISFCM clustering algorithm was applied to carry out the segmentation experiment of the NDT images. The results demonstrate that the proposed ISFCM clustering algorithm can not only effectively segment the NDT images with unbalanced grayscale distribution, but also extend its application scope, and enhance the robustness and adaptability.