基于改进Canny算子的脑部MR图像分割算法的研究

Brain MR Image Segmentation Algorithm Based on Improved Canny Operator

针对传统Canny算子在脑部核磁共振图像分割时存在的易受噪声干扰等问题,提出改进的Canny算子。首先将Gabor滤波器和Gaussian滤波器相结合,滤除MR图像中的伪影以及噪声点,然后进行非极值抑制确定边缘,最后根据Ostu思想,自动计算双阈值来去除假性边缘,得到最终分割后的脑部MR图像。实验结果证明,改进后的Canny算子能够有效地检测出对象边缘,并对脑部MR图像中常见噪声伪影具有很强的鲁棒性。

The traditional canny operator has the problem of being susceptible to noise interference in brain MR image segmentation.Considering these problems,an improved Canny operator was proposed.Firstly,Gabor filter and Gaussian filter were combined to filter out artifacts and noise points in brain MR images.Then the edge was determined by non-extreme value suppression.Finally,according to the Ostu idea,the double threshold was automatically calculated to remove the false edge,and the final segmented brain MR image was obtained.Experimental results showed that the improved Canny operator could effectively detect the edge of objects,and was robust to common noise artifacts in brain MR images.