去除红外图像中脉冲噪声的双边加权中值滤波

Bilateral Weighted Median Filter for Removing Impulse Noise in Infrared Images

针对现有方法在去除红外图像的脉冲噪声时,未能有效保持图像的边缘细节和纹理结构,提出了基于统计检测的双边加权中值滤波算法。算法根据脉冲噪声的取值和分布特征,用最小和最大像素值以及统计规律进行噪声检测;对检测出来的噪声像素,以多尺度的方式、自适应地用双边加权系数对邻域中的无噪像素和已经去噪处理的像素进行频次加权,然后取它们的中值作为当前噪声像素的估计值。其中双边加权系数自适应于距离邻近度与灰度相似度。实验结果表明,相对于部分现有方法,所提方法去噪所得的EPI和SSIM值更高,去噪图像的视觉效果更佳。

In view of the fact that the existing methods could not effectively maintain the edge details and texture structures of the image when removing impulse noise in infrared images,bilateral weighted median filter based on statistical detection is proposed.According to the value and distribution characteristics of impulse noise,the minimum and maximum pixel values and the statistical law are used to detect the noise.For the detected noisy pixels,in a multi-scale manner,the noise free pixels and the pixels that have been denoised in the neighborhood are weighted adaptively by the bilateral weighting coefficients,and then their median value is taken as the estimated value of the current noisy pixels,in which the bilateral weighting coefficients are adaptive to the distance proximity and intensity similarity.The experimental results show that compared with some existing methods,the proposed method achieves higher EPI and SSIM values,and the visual effect of its denoised images is better.