Abstract: In order to quickly remove the impulse noises in images with the texture details and edge structures retained, an image denoising algorithm based on trimmed mean and Gaussian weighted median is proposed.Based on the gray features and statistical features of impulse noises, the local statistics is used for noise detection, identifying the pixels with minimum or maximum gray values and with less correlation with neighboring pixels as the noisy pixels.Then the noisy pixels in the smooth regions and detail regions are processed respectively by the adaptive trimmed mean filtering and Gaussian weighted median filtering for noise reduction.Experimental results show that the proposed algorithm outperforms other algorithms in terms of visual effects, Peak Signal to Noise Ratio(PSNR), Structure SIMilarity(SSIM), and computational efficiency.It also preserves the texture details and edge structures of images better while eliminating the noises.

Key words: image denoising, impulse noise, noise detection, median filtering, trimmed mean filtering, Gaussian weighted median filtering

摘要： 为快速准确地滤除图像中的脉冲噪声并较好地保持图像的纹理细节和边缘结构，提出一种基于修剪均值与高斯加权中值滤波的图像去噪算法。根据脉冲噪声的灰度特征与统计特征，以局部统计方式进行噪声检测，将灰度取最小值或最大值且与邻域像素相关性较小的像素识别为噪声像素。对于图像平滑区域和细节区域中的噪声像素，使用自适应修剪均值和高斯加权中值滤波算法进行去噪处理。实验结果表明，该算法在视觉效果、峰值信噪比、结构相似性及计算速度上均优于对比算法，并且能够在彻底滤除噪声的同时，较好地保持图像的纹理细节和边缘结构。

关键词: 图像去噪, 脉冲噪声, 噪声检测, 中值滤波, 修剪均值滤波, 高斯加权中值滤波

CLC Number: 

• TP391