基于预估点扩展函数的光学遥感图像去模糊研究

Image Deblurring Study for Spaceborne Optical Remote Sensing Using Estimated Point Spread Function

为使航天相机获得的遥感图像的视觉质量更佳,提出了一种有效而稳定的基于估算点扩展函数(PSF)的遥感图像快速复原方法。先用改进的刃边法快速估算PSF尺寸和形状,用直接边缘检测法处理,由自适应阈值算法抑制噪声产生的波动,有效减小了PSF估算的存储与时间开销。再基于预估的PSF,用非盲反卷积复原算法复原图像,通过带权补偿去模糊算法由正则化反卷积和高频补偿,有效抑制源于PSF估算误差的振铃现象。根据视觉效果和客观评价指标对该复原方法与经典算法的效果进行了比较,结果表明:该法在存在噪声和点扩展函数估计误差的影响时仍有良好的鲁棒性,图像质量提高幅度大,速度快,可行性高,已成功用于浦江一号(PJ-1)卫星的遥感图像处理。

To obtain better visual effect of remote sensing images captured by spaceborne imagers,a novel onorbit deconvolutional restoration method based on point spread function(PSF)was presented in this paper.The improved edge method was used to estimate the size and shape of PSF.The direct edge method was used and the adaptive threshold algorithm was applied to suppress the fluctuation caused by noise,which reduced the memory and operation time for PSF estimation.The non-blind deconvolutional method was used to restore the images based on estimated PSF.The original image was compensated by the regular deconvolutional and high frequency compensation with weighted deblurring method,which would suppress the ringing effect caused by the error of PSF estimation.The comparison of the traditional algorithm and method proposed was carried out according to the visual impression and evaluation criterion.The results showed that the method proposed had good robustness when the noise and estimation error of PSF existed,and the image quality had improved greatly,which was fast and feasible.The method was successfully used in the image treatment of remote sensing in PJ-1satellite.