摘要
为了改善空间光学遥感图像的质量,提出了基于受限全变差正则化的图像去模糊方法。通过多通道盲反卷积法估算点扩展函数,从而将遥感图像的去模糊问题转化为非盲复原问题。然后通过快速梯度投影算法求解非平滑最优化问题,得到去模糊图像。对不可避免的点扩展函数估算误差和噪声,该方法不会引入明显的振铃和噪声放大。对全色遥感图像的实验结果表明,该方法在保持图像均值的同时,将拉普拉斯能量由11.1455提升至57.5541,去模糊后的图像相对原始图像的结构相似度指标为0.9824。直观效果与客观评价指标都表明该方法可以有效提升遥感图像质量。
In order to improve the quality of images captured by spaceborne optical remote sensors, a deblurring method based on constrained total-variation regularization is proposed. First of all, the deblurring problem is transformed into a non-blind one via estimation of the point spread function (PSF) using multichannel blind deconvolution. The deblurred image is obtained by applying fast gradient projection algorithm to this non-smooth optimization problem. On the inevitable existence of PSF estimation error and noise, the proposed method does not introduce significant ringing and noise amplification. Experimental results based on panchromatic remote sensing images show that it can preserve mean value and meantime increase the energy of Laplacian from 11. 1455 to 57. 5541. The structural similarity index between original and deblurred images is up to 0. 9824. Both visual effect and evaluation indicators demonstrate that the proposed method can effectively improve the quality of remote sensing images.
出处
《激光与光电子学进展》
CSCD
北大核心
2013年第11期88-93,共6页
Laser & Optoelectronics Progress
基金
吉林省科技发展计划项目(201000526)
关键词
图像处理
光学遥感
图像去模糊
全变差
正则化
点扩展函数
image processing
optical remote sensing
image deblurring
total-variation
regularization
point spread function