The neighborhood weight function built in image segmentation using the improved fuzzy c-means clustering algorithm fails to simultaneously consider space structure and grayscale range information, which results in the problem of noise sensitivity and rough dealing with edge texture information. To this problem, in this paper, a FCM algorithm combined with wavelet transform and improved neighborhood weights is proposed. First, the algorithm deals with the original gray image by using the adapt threshold denoising method, which is based on wavelet used for multi-resolution analysis. Second, it constructs an improved neighborhood weight function based on the local spatial neighborhood information and grayscale range information of the image patches by combining with the thought of bilateral filtering in the reconstructed image. The experiment results show that the proposed algorithm has a higher accuracy of segmentation than the traditional FCM algorithm and improved FCM algorithm and is more robustness to the strong noise with more smooth image edges.