提出基于视觉显著性特征的乳腺钼靶X射线肿块检测方法:首先从局部显著性、全局显著性和稀少性三方面计算显著图,利用显著图加权增强目标;然后根据前景目标数迭代确定分割阈值对加权后图像阈值分割;最后将分割后的前景区域视为疑似肿块区域,利用融合显著性特征及基于中心-轮廓距离的肿块形态特征识别肿块。本文利用MIAS数据库中多幅的乳腺X线图像进行实验验证,结果表明,本文提出的方法能够准确地分割肿块区域,肿块识别准确性较高。 A detection method of breast mammography based on visual saliency features was proposed. First, saliency maps were calculated from three aspects: local saliency, global saliency and sparseness. The saliency maps were weighted using saliency map weights. Then, the segmentation threshold Thresholding the weighted images. Finally, the foreground area after segmentation is regarded as the suspected tumor area, and the tumor is identified by using the fusion saliency features and the tumor shape features based on the center-outline distance. In this paper, many mammograms in MIAS database are used to verify the experimental results. The results show that the method proposed in this paper can accurately segment tumor mass and the accuracy of tumor mass identification is high.