双变量模型下的非下采样Contourlet变换图像去噪

Using the Bivariate Model for Image Denoising via Nonsubsampled Contourlet Transform

该研究利用非下采样Contourlet变换的平移不变性和多方向选择性,考虑变换域内子带系数尺度间和尺度内的双重相关性,自适应地调节双变量模型下子带系数的收缩量,使子带系数的收缩量与子带含有图像细节内容的多少成比例.仿真结果表明,与仅考虑子带系数尺度间相关性的去噪算法相比,该算法在去除噪声的同时能有效保持原图像中的细节和纹理信息,改善恢复图像的主观视觉效果,提高恢复图像的PSNR值.

This algorithm takes the advantage of translation-invariant and multidirection-selectivity caused by nonsubsampled contourlet transform, and exPloits the inter-scale and intra-scale correlations of coefficients, which adaptively shrink according to the information of subbands in the bivariate model. Compared with some denoising methods which are just based on the inter-scale correlations, the simulation results show that this algorithm obviously improves visual quality and Peak Signal-to-Noise Ratio （PSNR）, and effectively preserves the details and texture information of original images.