基于邻域信息的自适应中值滤波算法

针对图像中普遍存在的脉冲噪声，提出了一种自适应中值滤波算法，该算法在有效去除噪声的前提下能够保留更多的图像细节。首先，根据脉冲噪声灰度值为0或1的特点初步区分图像中的噪声点和信号点；其次，在每一个可疑噪声点周围取一定大小的邻域，通过判断该可疑噪声点与邻域内其他像素点之间相关性的大小进一步判断该点是否为真正噪声点，若为真正噪声点则利用邻域内所有可靠像素点的中值代替，否则输出原信号点。利用可见光及红外图像将所提算法与几种算法（如传统中值滤波算法、极值中值滤波算法，等）进行比较，实验结果表明该方法能够获得最高的峰值信噪比，去噪效果最佳。

Adaptive median filtering algorithm based on neighborhood correlation

Aiming at the impulse noise widely exiting in images, an adaptive median algorithm was proposed in this paper. The algorithm could suppress noise effectively and preserve more image details. Firstly, noise pixels and signal pixels were preliminary distinguished according to the characteristic of the impulse noise gray value of 0 or 1. Whether the pixel was noise or not depended on the correlativity of adjacent pixels in the window centered on the pixel. The gray value of contaminated pixel was reconstructed by the median of those uncontaminated pixels in the window. And the signal pointed directly output without changes. The results with visible-light images and infrared images show that the proposed method has a better performance because of the highest peak signal-to-noise ratio than other methods, such as Traditional Median (TM) and Extremal Median (EM) filtering algorithm.