基于边缘软判决的小波域自适应图像去噪

Adaptive image denoising based on edge soft-decision in wavelet domain

提出了一个新的图像去噪方法。该方法基于非抽样小波变换的多分辨分解,在各尺度下对小波系数进行了边缘和非边缘分类,并根据它们的不同统计特性运用了不同的估计技术。鉴于边缘分类的不确定性,提出了依概率的软分类技术,通过计算边缘发生的概率,判决当前系数应该采用哪一种估计。仿真结果表明:该方法在滤除图像噪声的同时,边缘得到了保持,较目前存在的一些方法更具有优越性。

A new image denoising method is proposed based on image multiresolution decomposition by undecimated wavelet transform. At each resolution level, the wavelet coefficients are classified into edge-related and non-edge-related sets, which are processed by different denoising techniques separately, based on different statistical models. Due to the uncertainty of edge decision, a soft-decision method is employed to decide which denoising technique can be applied to the wavelet coefficient by the probability of edge. The experiments show the image noise is filtered out, while the edges are preserved, and the proposed approach outperforms some existed denoising methods.