Abstract: An adaptive median filter that can remove impulse noise in medical images is implemented, and it is compared with traditional median filter and averaging filter by two objective evaluation criteria, root-mean-square error and destroyness degree of noise to original image. The pixel at the center of filtering window is detected to determine whether it is an impulse noise based on given condition. Median filter with adaptive window is applied to eliminate the noise. The algorithm overcomes the shortcoming of median filter that replaces the value of each pixel by the median of the gray levels in the neighborhood of that pixel, and reduces unnecessary loss of image detail. Simulation results based on MATLAB show that, for larger impulse noise density, it not only eliminates noise effectively, but also preserves the edges and detail information of the objects. The algorithm has been used in the virtual endoscope system, and the filtering performance is very satisfactory.

Key words: virtual endoscope system, medical image filtering, adaptive algorithm, median filter, impulse noise

摘要： 实现了一种滤除医学图像脉冲噪声的自适应中值滤波算法，用均方根误差和噪声对原图像的毁坏程度两个客观评价指标对该算法及传统均值、中值滤波方法进行了比较与评价。根据设定条件检测滤波窗口中心像素是否为脉冲噪声，采取滤波窗口自适应的算法来滤除脉冲噪声，去除了传统中值滤波对所有像素均用中值代替的弊端，减少了不必要的图像细节损失。基于MATLAB的仿真实验表明，对于较大密度的脉冲噪声，该算法在有效抑制噪声的同时，能较好地保护边缘和细节信息。该算法已应用于虚拟内窥镜系统中，取得了令人满意的效果。

关键词: 虚拟内窥镜系统, 医学图像去噪, 自适应算法, 中值滤波, 脉冲噪声