| 基于广义变分正则化的红外图像噪声抑制方法 |
| |
| 引用本文： | 钱伟新,王婉丽,祁双喜,程晋明,刘冬兵.基于广义变分正则化的红外图像噪声抑制方法[J].红外与激光工程,2014,43(1):67-71. |
| |
| 作者姓名： | 钱伟新 王婉丽 祁双喜 程晋明 刘冬兵 |
| |
| 作者单位： | 1.中国工程物理研究院流体物理研究所,四川绵阳621900 |
| |
| 基金项目： | 国家自然科学基金青年基金(61307014);中国工程物理研究院科学技术发展基金(2011B0402006);国家重大科学仪器设备开发专项(2011YQ130019);科工局技术基础科研项目(2012XXX) |
| |
| 摘 要： | 文中提出了一种广义变分正则化的红外图像噪声抑制方法，该方法采用p-范数代替目前广泛被采用的全变分范数作为正则项，构造了用于抑制图像噪声的展平泛函，从而将图像噪声抑制问题转化为能量泛函优化问题。通过推导，得到了相应的用于图像噪声抑制的非线性偏微分方程，并采用固定点迭代算法进行线性化求解，使得迭代解稳定收敛。数值试验结果表明，该方法能够有效地去除图像噪声，较之全变分图像噪声抑制方法，新方法进一步提高了对小宽度图像边缘的保持能力，是一种有效且性能优良的红外图像噪声抑制方法。
|
| 关 键 词： | 广义变分 正则化 全变分 非线性偏微分方程 |
| 收稿时间： | 2013-05-10 |
Generalized variation-based regularization method for infrared image denoising |
| |
| Affiliation: | 1.Institute of Liquid Physics,China Academy of Engineering Physics,Mianyang 621900,China |
| |
| Abstract: | A generalized variation (GV) regularization based infrared image denoising method was proposed in this paper. In the new method, a p-norm was used as regularized term to replace total variation (TV) norm in traditional TV based image denoising methods which were used popular in image processing domain. Then a smoothing functional was constructed for noised removal. Thus, the problem of image denoising was transformed to a problem of a functional minimization. A nonlinear partial differential equation (PDE) was deduced from the new image denoising model. To solve the nonlinear PDE, the fixed point iteration (FPI) scheme was introduced to linear the PDE. The stability and convergence of regularized solution were ensured by FPI scheme. The numerical experimental results show that comparison with TV regularized method, the GV regularized method can preserve image edge including those small width edges more efficiently while removing noise. The GV regularized method is an efficient image noise removed method with better performance of noise removal and edge preserving. |
| |
| Keywords: | |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《红外与激光工程》浏览原始摘要信息 |
|
点击此处可从《红外与激光工程》下载全文 |
|
