为了在去除红外图像的脉冲噪声的同时, 有效保持和恢复图像的边缘细节, 提出了基于灰度特征和众数原则的迭代双边中值滤波方法。此方法根据脉冲噪声的灰度特征以及众数原则, 将取最小和最大值、而在邻域的灰度分布上孤立的像素识别为噪声。根据基于空间距离和灰度相似的加权系数, 对邻域中的无噪像素与已经去噪恢复的像素进行频次加权, 用频次加权中值作为噪声像素的估计值。其中, 以迭代遍历的方式执行去噪处理, 充分利用前次遍历处理的结果, 以去除高密度噪声。实验数据证明, 此方法去噪所得的 PSNR和 EPI值以及视觉效果均优于现有方法, 具有更好的去噪性能。

In order to effectively maintain and restore the edges and details of infrared images while removing the impulse noise, an iterative bilateral median filter based on intensity features and mode principle is proposed. In this method, based on the intensity features of impulse noise and the mode principle, the pixels that take the minimum and maximum values and are isolated on the intensity distribution of the neighborhood are recognized as noisy pixels. According to the weighted coefficients with respect to the spatial distance and intensity similarity, the noiseless pixels in the neighborhood and the pixels that have been denoised and restored are weighted by the frequencies, and the frequency weighted median is used as the estimated value of noisy pixels. Furthermore, the denoising processing is performed in the way of iterative traversal processing, which makes the most of the results of the previous traversal processing to remove high density noise. The experimental data confirm that the PSNR and EPI values and the visual effects achieved by the proposed method are better than the existing methods, with better denoising performance.