摘要
提出了一种基于支持向量值和非抽样方向滤波器组的图像去噪算法。该算法通过构造支持向量值方向滤波器组(SVDFB)对噪声图像进行多尺度、多方向分解,同时考虑到分解系数服从广义高斯分布的统计特征,采用局部自适应贝叶斯阈值方法实现图像去噪。仿真结果和实验分析表明,该算法的峰值信噪比和去除噪声后图像的视觉效果都有明显提高,同时有效保留了原图像的纹理和细节信息。
In order to denoise image,this paper proposed a new method based on support vector value and undecimated directional filter bank.It used the proposed algorithm to decompose noise images at multi-scale and multi-direction by support vector value directional filter bank（SVDFB）.It was considered that the coefficients statistic characteristic of test image by SVDFB obeyed general Gaussian distribution,adopted local adaptive Bayes shrinkage factor.The simulation results and experimental analysis show that the proposed algorithm outperforms in both peak signal-to-noise ratio（PSNR） and visual quality,and effectively preserves texture and detail information of original images.
出处
《计算机应用研究》
CSCD
北大核心
2011年第6期2375-2377,2380,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60772121)
安徽省教育厅重点科研计划资助项目(KJ2010A021)
关键词
图像去噪
支持向量值
非抽样方向滤波器组
贝叶斯阈值
image denoising
support vector value
undecimated directional filter bank
Bayes threshold value
