A method based on visual saliency is proposed for mass detec tion in mammograms in this paper.Firstly,local saliency,global saliency and rarity sa liency are computed and fused,and the fusion saliency is used for enhance masses by weighting mammogram s.Secondly, calculate the segmentation threshold according to the number of foreground targe t,and segment the weighted image by the calculated threshold.Lastly,regarding foreground of the segmented image as suspected masses,and extract the characteristics of fusion saliency features a nd morphological characteristics based on center-boundary distance for the foreground target.Us e the characteristics to identify masses. We use mammograms from MIAS dataset to verify the proposed method.The result s show that this method can segment masses more accurately compared with other methods.This method can result in high recognition accuracy for detection masses.