To obtain a favorable segmented result of an image, a transition region method, based on the difference of local fuzzy variance between transition region and background, was proposed and used for image segmentation. First, the method to construct fuzzy variance based on fuzzy set was extended to the case of two dimensional image; second, a transition region was obtained by the local fuzzy variance; and at last the final segmented result was acquired by a multi-threshold selecting method. Experimental results show that the proposed method is robust and it can provide satisfactory segmented result.