提出了一种新的有效去除图像中吉布斯(Gibbs)噪声的方法。通过非下采样塔式滤波器(NSPF)对阈值去噪后的图像进行分解,利用全变差模型对分解图像分别建立去噪模型,重构图像,再利用重构图像和原始图像的残差并结合全变差模型得到细节补偿图像,最后把重构图像和补偿图像叠加得到最终去噪图像。实验结果表明,该方法可以有效地消除图像中的吉布斯伪影及噪声,在去噪图像峰值信噪比(PSNR)和边缘保持性能上都优于已有的算法。

A new algorithm for removing the Gibbs-like artifacts in image denoising is presented. The original image is decomposed by using the nonsubsampled pyramid filter (NSPF) and the decomposed image model is built based on total variation model. Then the preliminary denoised image is produced by reconstruction. A detail compensation image is further obtained by using the difference between the preliminary denoised image and the original image. Finally,the denoised image can be obtained by adding the compensation image to the reconstructed image. Experiments show that the proposed scheme can remove Gibbs-like artifacts and image noise effectively. This algorithm outperforms the existing schemes in regard of both the peak-signal-to-noise-ratio (PSNR) and the edge preservation ability.