基于形态学多尺度修正的脑水肿区域分割

脑CT图像组织结构较为复杂且灰度不均匀,直接采用分水岭分割会导致较严重的过分割,采用阈值标记控制的分水岭分割通过限定可分割区域,可以较好地减轻过分割,但容易出现目标标记不准确的问题。为此提出了一种基于形态学多尺度修正的标记控制分水岭分割方法,首先对原始图像进行线性拉伸和指数增强,提高水肿与正常区域的对比度;然后在形态学梯度图像基础上,根据不同像素梯度值确定结构元素的大小,对图像进行形态学多尺度修正,以消除局部极小区域,保证修正过程中目标轮廓不发生较大偏移;最后采用标记控制的分水岭变换对图像进行分割。实验结果表明,该方法可对脑部水肿区域进行较精确的分割。

Brain edema segmentation based on morphological multi-scale modification

Brain CT image structure is very complicated,moreover,very uneven in gray scale.Implementing direct watershed segmentation may lead to more serious over-segmentation.Using threshold marker controlled watershed segmentation may restrain the over-segmentation problem to a considerable extent.However,it is prone to problems of inaccurate target marking.In this paper,we propose a marker-controlled watershed segmentation method based on morphological multi-scale modification.We begin by linear stretching the original image to enhance its indexes.It enhances the contrast of the brain edema and normal area.Then based on the morphological gradient images,determine the scale of the structural elements according to the different pixel gradient values.The morphological multi-scale image correction eliminates the local minimum area,thus guarantees that the target contour or outline correction process does not produce major displacements.Finally,marker-controlled watershed image segmentation is then effected.Experiments show that this method has a higher accuracy of segmentation of edema region in brain CT images.