基于权重自适应形态学的周期性噪声去除方法

A Periodic Noise Elimination Algorithm Based on Morphological Filtering with Auto-adapted Weights

针对去除周期性噪声的同时容易造成图像的失真或降噪效果不理想的问题,提出了一种基于权重自适应形态学的周期性噪声去除方法。该方法使用不同尺度的结构元素对图像的周期性噪声进行串行处理,再将串行处理的结果并行处理,并通过自适应权值算法来构建复合级联滤波器,使用该滤波器滤除图像的周期性噪声。为了验证该算法的去噪性能,对周期性噪声及混合噪声进行了常用去噪算法的对比性实验。结果表明,视觉上,使用该算法去噪后的图像去噪效果较好且图像边缘和细节比较清晰;定量评价标准上,使用该算法去噪后的图像的PSNR和SSIM都较高。因此,该算法有效地抑制了图像中的周期性噪声,同时较好地保持了图像的几何特征,具有更好的鲁棒性。

Aiming at the problem that it is easy to cause the image distortion or poor noise reduction while eliminating the period noise, we propose a periodic dennising method based on morphological filtering with auto-adapted weights. In this method,periodic noise of im- ages is processed serially by structural elements of different scales, then the results of serial processing are processed in parallel by con- structing composite cascade filter using multi-structural elements. In order to verify the denoising performance of the proposed algorithm, some denoising algorithms are used to eliminate periodic noise and mixed noise. The experiments show that the de-noised image obtained by the proposed algorithm is less residual noise and clearer textures than other algorithms visually. At the same time, in the quantitative evaluation standard,the PSNR and SSIM of de-noised image obtained by the proposed algorithm are higher. So,it is robust,not only effectively restraining periodic and mixed noise,but also preferably maintaining image geometry characteristic.