Abstract: A novel image denoising method based on region segmentation is proposed in this paper.Hard thresholding with linear shift invariant Discrete Wavelet Transform（DWT） and NeighShrink with Stein’s Unbiased Risk Estimate（SURE） are most important and efficient methods for image denoising.The former is much fitter than the later for the smooth image denoising，and it is opposite for the texture.The proposed algorithm based on the two methods.Firstly，the noisy image is denoised by hard thresholding with linear shift invariant DWT and by NeighShrink with SURE respectively，then two denoised images are obtained.Secondly，the noisy image is divided to be smooth region，non-smooth region and transitional region.The final denoising image is that the three regions are respectively determined by the two denoised images with the certain proportional divisor.The experimental result shows that the proposed algorithm can absorb the advantages of the two methods，a denoised image，with higher PSNR，rich texture details and better vision and quality，is obtained.

摘要：

提出一种基于区域分割的图像去噪方法。该方法利用具有平移不变性的DWT去噪法和NeighShrink_ SURE去噪法对平滑图像和纹理图像分别具有良好去噪效果，遂将含噪图像进行区域分割得到平滑、突变和过渡三个区域，最终去噪图像的三个区域分别由两种方法得到的去噪图像加权来确定。实验结果显示，该方法利用了前两种算法的优点，得到了具有较高峰值信噪比、较完整保留图像细节而且具有更佳视觉效果的去噪图像。