随着光谱分辨率越来越高，高光谱图像更容易受到噪声的干扰，直接用传统的检测算子会产生较高的虚警。针对RX 算法存在较大噪声干扰的问题，提出了一种基于混合噪声评估的RX 异常检测方法。首先对高光谱图像进行分块，利用滤波的思想选取均匀图像块；考虑图像光谱-空间信息，运用多元线性回归分析对均匀图像块进行混合噪声评估；然后将高光谱图像和混合噪声进行作差，消除噪声的干扰；最后运用RX 算子进行异常检测。实验结果表明，该方法达到了消除噪声的效果，与RX 和MNF-RX 算法相比具有更好的目标检测性能。

With a higher spectral resolution, hyperspectral images are more susceptible to noise; furthermore, conventional detection operators generate a high rate of false alarms. Aiming to overcome the problem of large noise interference as related to the RX algorithm, a new method of RX anomaly detection based on mixed noise is proposed. First, a hyperspectral image is divided into blocks, and a uniform block is selected via filtering. The mixed noise is estimated by performing multiple linear regression analysis, which considers spectral and spatial information, on the uniform image block. Then, the estimated mixed noise is subtracted from the hyperspectral image to eliminate noise interference. Finally, anomaly detection is performed by implementing the RX algorithm. The experimental results showed that the proposed method effectively eliminates noise and achieves better detection performance than that of RX and MNF-RX algorithms.