基于平稳小波/Contourlet域非线性扩散的印章图像去噪

Seal Image De-noising Based on Stationary Wavelet/Contourlet Transform and Nonlinear Diffusion

在印章鉴别系统中,正确鉴别印章的前提条件是得到高质量的预处理图像.提出了分别将平稳小波变换、Contourlet变换与各向异性扩散相结合的印章图像去噪方法.首先对图像进行平稳小波/Contourlet分解,然后高频部分和低频部分分别采用自适应对比度扩散和全变差扩散,最后重构图像.给出了实验结果,并与现有的小波阈值收缩和全变差扩散结合的方法、基于改进的Contourlet变换的自适应对比度扩散方法的图像去噪效果进行了主观视觉上的比较,同时也依据均方差(MSE)、峰值信噪比(PSNR)等评价指标作了定量分析,且对比了各算法的运行时间.结果表明,本文提出的方法去噪效果更为优越:不但抑制噪声的能力更强,而且能够更好地保留印章图像原有的边缘特征.

In the seal imprint verification system, obtaining high quality preprocessed image is the prerequi- site for correctly verifying the seal imprint of documents. A seal image de-noising method combining station- ary wavelet]contourlet transform with nonlinear diffusion was proposed. Firstly, an image was decomposed by stationary wavelet]contourlet transform. Then adaptive-contrast-factor diffusion and total variation diffusion were applied to high-frequency components and low-frequency components, respectively. Finally the image was synthesized. The experimental results were given. Comparisons of the image de-noising results were made with those of the image de-noising methods based on the combination of wavelet shrinkage with total variation diffusion, and the combination of improved contourlet transform with adaptive-contrast-factor diffusion. Run- ning time of the above-mentioned algorithms was also contrasted. It is shown that the proposed image de- noising methods based on stationary wavelet/contourlet transform and nonlinear diffusion can obtain superior results. They can both remove noise and preserve the original edges more efficiently.