中文摘要：

针对脑部核磁共振成像（magnetic resonance imaging,MRI）图像中因噪声、灰度不均匀及边界模糊不连续等造成肿瘤难以准确分割的问题,提出一种基于形态学重建和梯度分层多尺度修正的分水岭分割方法。首先对原始图像进行形态学混合开闭重建以平滑去噪,同时保留目标轮廓信息;然后根据梯度图像的三维地貌体积对其进行分层多尺度修正,自适应地确定修正所需的结构元素尺寸,对低梯度层级采用较大尺寸结构元素进行闭运算修正,消除产生过分割的非规则局部极小值,而对较高梯度层级则采用较小尺寸的结构元素,保持区域轮廓的位置不变;最后在修正基础上,运用标准分水岭变换实现图像分割。实验结果表明,该方法与标准分割的相似度指数和Jaccard指数均较高,且过分割率和欠分割率均较低,具有较好的分割效果。

英文摘要：

The accurate segmentation of tumor in brain MRI images is usually difficult due to noise, gray inhomogeneity, fuzzy and discontinuous boundaries. For the purpose to get precise segmentation with less contour position bias, this paper presented a novel watershed algorithm based on morphological reconstruction and gradient layered modification. Firstly, it employed marphological hybrid opening and closing by reconstruction operators to smooth and denoise the original image, while retaining the target contour information. Then, it stratified the gradient image by the volume of three-dimension landform, further modified the lower gradient layers with large-sized structuring elements, whereas the smaller-sized to the higher layers. Thus it removed most local minimums caused by irregular details and noises, while region contour positions corresponded to the target area. Finally, it employed morphological watershed algorithm to implement segmentation on the basis of multi-scale modified image. The experimental results show that the suggested method can achieve more accurate segmentation result, owing to its lower over-segmentation and under-segmentation, as well as the higher similarity index compared with the standard segmentation.