棉花噪声图像的检测与滤波

Detection and filtering of cotton noise image

针对当前图像噪声检测中存在的“虚噪声”以及中值滤波对图像边缘特征的影响,提出将图像饱和度通道像素中极大值或极小值的点列为可疑噪声点,在图像色调通道检测出噪声点与色相相近可疑噪声点,对明度通道色相相近的可疑噪声点进行再判别,筛选出噪声和有用信号。同时,对选取窗口内4个特定方向上的像素点灰度值进行排序,对排序后的各个中值进行加权计算,用计算结果替代窗口中心像素的灰度值。试验结果表明,该方法能够有效检测彩色图像饱和度通道表现出的“虚噪声”,在去除噪声的同时有效保护图像细节。与其他滤波方法相比,峰值信噪比(PSNR)平均提高了8.2%,归一化均方误差(NMSE)至少降低了0.004,在滤波性能与滤波速度上都具有优势。

In view of the“virtual noise”in current image noise detection and the impact of median filtering on the image edge features,it is proposed to list the points with the maximum or minimum values in the saturation channel pixels of the image as suspicious noise points.The noise points are detected as suspicious noise points in the color channel of the image,and suspicious noise points with similar color in the brightness channel are identified to filter out noise and useful signals.At the same time,sort the grayscale values of pixel points in four specific directions within the selected window,calculate the weighted values of each sorted median,and replace the grayscale values of the central pixel in the window with the calculated results.The experimental results show that this method can effectively detect the“virtual noise”exhibited by the saturation channel of color images,effectively protecting image details while removing noise.Compared with other filtering methods,the peak signal-to-noise ratio(PSNR)has been improved by an average of 8.2%,and the normalized mean square error(NMSE)has been reduced by at least 0.004.It has advantages in filtering performance and filtering speed.