基于离群特征的γ辐射图像去噪方法

针对Co60辐射环境中γ光子穿透CMOS图像传感器时致使场景图像存在斑块噪声的问题，提出了一种基于离群特征的γ辐射图像去噪方法。首先在序列图像中逐点获取对应的像素序列，并将该像素序列进行光照归一化以消除图像帧之间光照差异影响；然后在光照归一化后像素序列中利用噪声像素值的离群特性判断当前像素点是否为噪点；最后利用序列中各点的一、二阶离群特征筛选有效像素序列，并将其均值进行逆光照归一化以作为噪点修复的像素值。所提方法与多种典型去噪方法分别在高剂量率区和低剂量率区的真实γ辐射图像上进行了对比实验，该方法均取得了最佳去噪效果。

Denoising method of gamma radiation image based on outlier features

In the Co60 radiation environment, gamma photons will cause speckle noise in the scene images when they penetrate the CMOS image sensor. So this paper proposed a denoising method for gamma radiation image, which was based on the outlier features of noise. This method firstly obtained a set of corresponding pixels in the sequence of images, and normalized the illumination of obtained pixels to eliminate the influence of the brightness difference among images. Then, it used the outlier features of the normalized pixels to detect the noise. Finally, it applied the first-order filtering and second-order filtering method to select non-noise pixels, and the noise-repaired pixel was the mean of non-noise pixel. It compared proposed method with various typical denoising algorithms on real gamma radiation images which captured in high-dose and low-dose scenes respectively, and it achieves best denoising effect according to the experimental results.