基于平移不变小波阈值算法的噪声图象估计

在小波基下简单的阈值算法几乎是最优的非线性估计, 若信号是分片光滑函数, 小波阈值方法所得的风险就接近于最佳值. 采用标准的小波阈值估计可较大幅度的提高PSNR, 但由于图像的边缘被Gibbs类的振荡所影响, 图像视觉质量较差. 本文采用改进的平移不变阈值算法可获得更高的PSNR, 更重要的是该方法明显地改善了图像的视觉质量.

Noisy Image Estimation With Translation Invariant Wavelet Thresholding

The simple thresholding algorithm in a wavelet basis is almost optimum nonlinear estimation. For piecewise smooth signal, the risk with wavelet thresholding is nearly the best value. The PSNR of Estimation with standard wavelet thresholding is enhanced, but edges of the picture are influenced by Gibbs′s surge, so the visual quality is lower. With a advanced translation invariant wavelet thresholding algorithm, the higher PSNR of Estimation is obtained, and it is more important to improve the visual quality of Estimation.