| 摘要: |
| 针对统计非局部均值滤波损坏图像的细节与鲁棒性双边带滤波去噪不充分的缺点，提出了一种基于统计非局部均值滤波与鲁棒性双边带滤波相结合的复合滤波算法。该复合滤波算法通过统计非局部均值滤波与鲁棒性双边带滤波线性组合，利用Stein无偏风险估计对复合算法中的参数进行估计。实验中，从主观与客观方面进行对比分析，证明所提出的复合算法体现了非局部均值滤波与双边带滤波的优点，能有效地去除噪声并更好地保留图像的细节信息，峰值信噪比提高1～2 dB。 |
| 关键词: 图像去噪 非局部均值 双边带滤波 Stein无偏风险估计 |
| DOI： |
|
| 基金项目:四川省教育厅项目(14ZB0211;14ZA0202)；人工智能四川省重点实验室开放基金项目（2015RZY01） |
|
| A hybrid image denoising algorithm based on Stein′s unbiased risk estimation |
| YANG Pingxian,HUANG Kunchao,ZHOU Bing |
| () |
| Abstract: |
| For the problems that the statistical property non-local means(SPNLM) destroys the image′s detail and the robust bilateral filter(RBF) can not denoise effectively，a new hybrid algorithm based on SPNLM and RBF is presented. By calculating a linear combination of SPNLM and RBF，the new algorithm uses Stein′s unbiased risk(SURE) to estimate the optimal parameter. The new hybrid algorithm is comparatively analyzed from the subjective and objective aspects.Experimental results indicate that the new hybrid algorithm takes the advantages of SPNLM and RBF,removes the noise more effectively and preserves more image details，and the peak signal-to-noise ratio(PSNR) of denoising image is increased 1～2 dB. |
| Key words: image denoising non-local means(NLM) bilateral filter Stein′s unbiased risk estimation |
