基于曲波变换和小波变换的图像去噪算法

Image Denoising Method Based on Curvelet and Wavelet Transform

针对应用曲波变换进行图像处理过程中所产生的伪影现象,提出一种基于快速离散曲波变换和小波变换的联合去噪算法。对噪声图片分别采用非抽样小波变换和快速离散曲波变换进行去噪,并对经过快速离散曲波变换去噪后的图像进行四叉树分解,根据分解结果对图像进行重构得到最终融合图像。实验结果表明,联合去噪算法对图像的边缘和均匀区域都有较好的去噪效果,能够有效抑制伪影,具有较高的峰值信噪比。

According to the artifacts that appear in the result images after applying curvelet denoising approach, a new method for denoising images corrupted with additive white Gaussian noise is proposed. Using Undecimated Wavelet Transform（UDWT） and Fast Discrete Curvelet Transform（FDCT） respectively for coding noisy image, applying quadtree decomposition to the result of FDCT approach and combining the attributes of both transforms by analyzing the result of quadtree decomposition, it is possible to get better effect. Experimental result shows that the Combined Denoising Method（CDM） denoises effectively in both homogeneous areas and areas with edges, suppresses artifacts evidently and has high peak signal-to-noise ratio.