结合非下采样轮廓波变换和混合阶次图像扩散的图像去噪

Image Denoising Using Hybrid Image Diffusion and Nonsubsampled Contourlet Transform

通过研究非下采样轮廓波变换理论及其在图像变换中的优点,提出一种新的基于非下采样轮廓波变换的图像去噪方法.该方法首先通过非下采样金字塔分解和非下采样方向滤波器组对待去噪图像进行非下采样轮廓波变换,然后采取不同阶次的图像扩散去噪算法分别对高频部分和低频部分进行去噪处理,最后将经过处理后的系数进行非下采样轮廓波逆变换便可得到去噪后的图像.通过实验结果表明,该方法不仅能有效的去除噪声,而且可以很好地保持边缘信息,整体性能优于近年来一些常见的去噪算法.

In this paper, the theoretics of nonsubsampled contourlet transform （NSCT） are studied, as well as its advantages in image transformation. A new algorithm based on nonsubsampled contourlet transform for image denoising is proposed. Firstly, the frequen- cy band on different scale and directon of the original image are acquired by using nonsubsampled directional filter bank and nonsub- sampled pyramid filter bank of NSCT. Secondly, different order of image diffusion algorithms were taken to process the high-frequen- cy and low-frequency parts respectively. Finally, denoised image was obtained by inverse contourlet transformed of these processed coefficients. Experimental results show that the algorithm can not only eliminate the noise effectively, but also preserve edge informa- tion very well. Its overall performance is superior to other denoising algorithms.