论文部分内容阅读
基于软阈值的广义交叉验证准则(GCV)已运用于图像去噪中,但此方法对图像峰值信噪比(PSNR)的改善有限,不能有效保持细节。针对此问题,改进了广义交叉验证准则,且将平移不变应用于去噪。即先将原图像分别进行水平和垂直方向上的平移,然后将平移后的每幅图像变换到小波域,使用分块广义交叉验证准则对小波系数去噪后实行小波逆变换和反平移。最后将反平移后的各图像平均,即得恢复的图像。实验结果表明,该方法能有效恢复图像细节,图像的PSNR值和主观效果都有较大改善。
Generalized Cross-Validation Criterion (GCV) based on soft threshold has been used in image denoising. However, this method can not improve the PSNR and can not effectively preserve the details. In order to solve this problem, generalized cross-validation criterion is improved, and invariance of translation is applied to denoising. That is to say, the original image is transformed horizontally and vertically respectively, and then each image after translation is transformed into the wavelet domain. The denoised wavelet coefficients are denoised by the block generalized cross-validation criterion and the inverse wavelet transform and anti-migration are performed. Finally, the anti-panned images are averaged to obtain the recovered images. Experimental results show that this method can effectively restore the image details, PSNR values and subjective effects of images have greatly improved.