The traditional fuzzy clustering algorithm (FCM) has the problem of uncertain initial cluster centers, and the gray and spatial information between pixels is not fully considered in image segmentation. In order to solve the above problem, a new fuzzy clustering image segmentation algorithm is proposed based on new distance matrix variance. The pixels are used to generate an improved new distance matrix, and the initial cluster center is selected according to characteristics of the new distance matrix. The number of cluster categories is determined combined with the variance, and part of noise is eliminated. The effectiveness determination is carried out on clustering result, and the best segmentation results are determined. Compared with the contrast algorithms, the average accuracy of proposed algorithm is increased by 4.55%. Experimental results show that the proposed method can effectively improve the average accuracy of image segmentation, and has a better effect on noise treatment.