基于L_0二阶偏导最小化和核回归模型的图像去噪方法  被引量:2

Image denoising via L_0 two-order partial derivative minimization and kernel regression

在线阅读下载全文

作  者:温云磊 王元全[1] 王怀彬[1] 

机构地区:[1]天津理工大学计算机与通信工程学院,天津300384

出  处:《天津理工大学学报》2015年第3期16-21,共6页Journal of Tianjin University of Technology

基  金:天津市自然科学基金重点项目(11JCZDJC15600)

摘  要:L0梯度最小化模型(LGM)作为一个最基本的数学工具已经成功的被用在了图像平滑领域.该模型最大的优势就是在处理图像的同时能够很好的保护图像的显著边缘.然而,作为总变差模型(TV)的改进版本,L0梯度最小化模型处理得到的结果图中却存在着比总变差模型更严重的阶梯效应并且不能够很好地保护图像的纹理和细节特征.为了克服这些缺点,本文提出将L0梯度最小化模型中的一阶导数推广到二阶偏导并且引入一个保真项,然后将其应用在图像去噪中.这个保真项是使用控制核作为核函数的移动最小二乘,即核回归模型.该模型虽然能够很好地保护图像的纹理特征,但是该模型处理得到的结果图中不仅会有流式效应并且不能很好的保护图像边缘.因此,本文利用二者的优势将其结合进行图像去噪.大量的实验结果表明提出的模型不仅具有良好的去噪属性并且在去除噪声的同时能够很好地保护图像的边缘和纹理特征.The L0 gradient minimization (LGM) method has been proposed for image smoothing very recently. As an improve- ment of the total variation (TV) model which employs the LI norm of the gradient, the LGM model yields much better results for the piecewise constant image. However, just as the TV model, the LGM model also suffers, even more seriously, from the staircasing effect and the inefficiency in preserving the texture in image. In order to overcome these drawbacks, in this paper, we extend the LGM model to I4 two-order partial derivative minimization and introduce an effective fidelity term into the 14 two-order partial derivative minimization. The fidelity term is an exemplar of the moving least square method using steering kernel, that is kernel regression (KR). This method can preserve the texture better, but it does not only have the disadvantage of leading to the serious flow-like effects but also often blurs the image edges. So this paper uses the advantages of both meth- ods fully and proposes a new image denoising algorithm. Enough experiments demonstrate that the proposed model does not only has better performance but also preserve the edges and texture well.

关 键 词:核回归 图像去噪 L0梯度最小化 L0二阶偏导最小化 

分 类 号:TP391.41[自动化与计算机技术—计算机应用技术]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象