光学 精密工程, 2012, 20 (9): 2078, 网络出版: 2012-10-12   

基于非负支撑域受限递归逆滤波的自适应图像盲复原

Adaptive blind image restoration based on NAS-RIF algorithm
作者单位
1 中国科学院 长春光学精密机械与物理研究所, 吉林 长春 130033
2 中国科学院 研究生院, 北京 100039
摘要
针对原始非负支撑域受限递归逆滤波(NAS-RIF)算法存在的缺点, 提出了一种自适应的NAS-RIF图像盲复原算法。首先, 在NAS-RIF算法的代价函数中加入正则化约束项和空域加权因子, 通过自适应地调整正则化参数和空域加权因子来改善算法的抗噪性能, 并确保复原的逼真和平滑。然后, 在算法的每次迭代中, 采用图像分割技术找到准确的目标支持域, 并用背景的平均值取代非均匀背景。最后, 利用N步重置共轭梯度法优化代价函数, 加快了算法的收敛速度。在不同信噪比条件下对两种模糊图像进行了实验, 结果显示, 采用本文算法得到的信噪比增益(ΔSNR)分别为6.315 3 dB和8.910 6 dB, 表明该算法具有较好的噪声抑制和边缘细节恢复效果。对低信噪比的退化图像, 本文算法也能得到更好的复原结果。
Abstract
To overcome disadvantages of the original Non-negativity and Support constraint Recursive Inverse Filtering (NAS-RIF) algorithm, an adaptive algorithm for the blind image restoration based on NAS-RIF algorithm was proposed. Firstly, regularization terms and space weights were added to the cost function of the original NAS-RIF algorithm. Through adaptively modulating the regularization parameters and space weights, not only the noise resistance ability could be improved, but the restored image could be smoothed. Then, image segmentation technique was employed in each iteration to find the precise object support region, meanwhile, the non-uniform background was replaced by the average background. Finally, the N-step-restart conjugate gradient routine was applied to optimization of the cost function, and then the convergence rate was enhanced. The experiments on degraded images derived from two kinds of blur operators were performed under different SNR (Signal Noise Ratio) conditions, and the ΔSNRs by proposed algorithm are 6.315 3 dB and 8.910 6 dB, respectively. The experiment results demonstrate that the proposed algorithm has a positive improvement in both reducing noises and preserving edges. Particularly, the proposed algorithm can obtain a better restoration result under a low SNR condition.

黄德天, 吴志勇. 基于非负支撑域受限递归逆滤波的自适应图像盲复原[J]. 光学 精密工程, 2012, 20(9): 2078. HUANG De-tian, WU Zhi-yong. Adaptive blind image restoration based on NAS-RIF algorithm[J]. Optics and Precision Engineering, 2012, 20(9): 2078.

本文已被 3 篇论文引用
被引统计数据来源于中国光学期刊网
引用该论文: TXT   |   EndNote

相关论文

加载中...

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!