提出了一种基于均值滤波的计算鬼成像（CGI）去噪方法，可以有效降低来自复杂环境的噪声干扰，提高CGI的成像质量。以尺寸为3×3的模板均值滤波器为例，设计了9组与均值滤波器相关的Hadamard移动散斑，并将各移动散斑依次照射到待测物体上，获得对应的桶探测器值。将9组桶探测值的累加值与散斑进行二阶关联，可获得被测物体去噪后的像。仿真和实验结果表明：在相同高斯和椒盐噪声环境下，所提方法获取的成像结果明显优于传统CGI的结果；所提方法具有较好的去噪能力，在不断变化的复杂环境中，有一定的应用优势。另外，所提方法将图像去噪中的均值滤波概念引入到关联成像中，提供了一种将信号处理方法用于CGI的新思路。

This paper proposes a computational ghost imaging (CGI) denoising method based on mean filtering to reduce noise interference from complex environments and improve the imaging quality of CGI. With a 3×3 template mean filter as an example, the paper designs nine groups of Hadamard shifted speckles related to the mean filter, illuminates the measured object by these shifted speckles successively, and obtains corresponding results by a bucket detector. After performing a second-order correlation on the speckles and the sum of the nine groups of values by the bucket detector, the denoised image of the measured object can be obtained. The simulation and experimental results show that compared with traditional CGI, the proposed method has better performance in improving the imaging quality under the same Gaussian and salt and pepper noises. Furthermore, it has a positive denoising effect and can be well applied in varying complex environments. In addition, the proposed method introduces the concept of mean filtering in image denoising to ghost imaging and provides a new idea for applying the signal processing method in CGI.