中文摘要：

针对无人机在视觉着陆过程中获取的图像受噪声影响的问题，设计了一种改进的中值滤波器。首先将物理学中的引力概念引入到图像像素间；其次给出了像素间的引力大小的数学表达式；最后通过像素间的引力来确定自适应窗口并进行相应的滤波处理。实验结果表明，与传统中值滤波以及开关中值滤波算法进行比较，该算法在图像去噪和细节保留方面有了很大的改善。

英文摘要：

In the process of UAV（ unmanned aerial vehicle）visual landing, noise may distort the landing image. To overcome the above deficiency, this paper proposed a novel improved median filter. Firstly, it introduced the concept of gravitation in physics into pixels. Secondly, proposed the mathematical formula of pixel gravitation. Finally, it got adaptive window according to pixel gravitation and processed the corrupted image by proposed algorithm. In experiment,it compared the proposed algorithm with two typical methods, named standard median filter and standard switching filter respectively. The results show that the pro- posed algorithm has better performance for capabilities of noise removal and detail preservation, especially effective for the ca- ses when the images are highly corrupted.