前景与背景分离的直方图重调整乳腺图像增强的新方法

提出了前景与背景分离的直方图重调整乳腺图像增强的新方法。首先利用高阶统计特征,并结合数学形态学方法检测出钙化点的感兴趣区域,得到了可靠性较高的前景图像,剩余部分设定为背景图像;然后分别对前景和背景的直方图进行重调整,得到了对比度增强的钙化点图像。通过与常规的增强方法进行比较发现,所提出的方法目标明确,能够较好的抑制非目标区域的影响。该方法在支持向量学习钙化点检测的预处理中得到了应用,较好的检测性能证实了算法的合理性。

Mammography image enhancement based on foreground and background redistributed histograms

In order to overcome the defect of existing ambiguous target in conventional method,a novel mammary gland image enhancing method is presented based on the foreground and background redistributed histograms.The suspect calcification region of interest(ROI) was first detected by using higher order statistic feature and morphology operator,so the ROI was defined as foreground image,and the rest was background image.Then a histogram redistribution scheme was adapted to process the foreground and background section separately.Compared with general enhancing methods,it can process image more clearly in region of target interest.The good performance in preprocess step of support vector machine microcalcification detection system proves that this method is valid and feasible.