摘要
利用计算机数值模拟的方法 ,对等功率谱滤波方法的图像恢复效果进行测定 .原始清晰图像选用以 (0 ,0 )为圆心的台阶式灰度图 ,且台阶灰度值非常接近 .这些清晰图像首先进行系统退化 ,然后加入噪声 (本文选用白噪声 )生成脏图 .图像恢复采用等功率谱滤波器 ,并将信噪比简化为常数k .图像恢复效果的好坏 ,用恢复后图像的信噪比和台阶的陡度 (台阶的下底宽减上底宽 ,单位为像素个数 )来定量测量 ,并结合普通图像的恢复 ,进行定性的视觉观察 .主要测量退化图像的信噪比对图像恢复效果的影响 ,结果是 :当信噪比 >1 5∶1时 ,图像恢复效果较好 ;当信噪比小于该值时 ,图像恢复效果较差 ;当信噪比下降到 5∶1时 ,图像恢复效果极差 ,几乎无法辨认 .其次是系统传递函数H对图像恢复效果的影响 ,也进行了一些讨论 .最后 ,通过对实际的天文图像进行恢复 ,验证了测量过程的合理性和结果的正确性 ,并列举了等功率谱滤波器的一个典型应用 。
The effect of equivalent power filtering in image restoration is analyzed via simulated data. The original clear image is a gray one which consists of a series of rings centered at (0,0). For different rings, the gray values are changes stepwise and smoothly. In order to obtain a dirty image, the original one is first transformed via a linear space invariance system. Then, white noise is mixed to the image. The Equal Power Filter is adopted for the image restoration and the SNR is supposed to be constant. SNR and the defined gradient of the steps, in cooperation with the visual effect, are accepted to characterize the performance of image restoration. Investigating the quality of the restoration affected by the SNR, a conclusive result is thus obtained: a satisfactory restoration can be realized when the SNR>15, while the effect of restoration becomes worse for the SNR<15. In the condition of SNR<5, the image is then completely not restorable. In addition, the restoration affected by the transfer function of the system is discussed in the paper. The present model of the Equal Power Spectrum Filter is confirmed by applying it to the observed image of objects. Finally, a typical application of the Equal Power Spectrum Filter is given.
出处
《天文学报》
CSCD
北大核心
2001年第2期206-214,共9页
Acta Astronomica Sinica
