一种频率域解混叠的CCD图像超分辨重建方法  

Kind of reconstruction method of CCD aliasing image in frequency domain

作　　者：赵秀影[1,2] 苏耘[2] 王敬梅 徐正平[1,4] 翟林培[1] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,长春130033 [2]空军航空大学,长春130022 [3]青岛城市建设投资(集团)有限责任公司,山东青岛266071 [4]中国科学院研究生院,北京100039

出　　处：《计算机应用研究》2010年第1期368-370,373,共4页Application Research of Computers

基　　金：国家"863"计划资助项目(2007AA701203)

摘　　要：在不增大光学系统的焦距,也不缩小CCD器件阵元的宽度的情况下,为了获得高分辨率的图像,对多幅具有互补信息的低分辨率图像进行研究。首先,把多幅具有互补信息的低分辨率图像通过傅里叶变换到频率域;接着,对频率域中高分辨与低分辨率图像之间关系进行建模,进而得到它们之间的关系;然后,通过对模型求解,得到高分辨率图像的频率值;最后,进行傅里叶反变换即可得到高分辨率图像。实验结果表明,该重建方法获得的高分辨率影像在视觉上比用任意一帧影像经过双线性插值获得的影像要清晰得多,与实际摄取高一倍的分辨率影像几乎达到一样。对bridge图像的恢复峰值信噪比可以达到34.767 3 dB。该方法是切实可行的,且恢复图像边缘视觉效果好。At the case of not increasing the focal length of the optical system, and not narrowing the array element width of CCD devices, in order to reconstruct super-resolution images ,researched multi-frames of under-sampled low-resolution images with the complementary information. First, they transformed multl-frames of under-sampled low-resolution images with the complementary information to the frequency domain by the Fourier transformation. Secondly, in the frequency domain, modeled the relationship between super-resolution image and the low-resolution images, which could get the relationship between them. Thirdly,acquired the frequency values of super-resolution images by solving the model. Finally, got super-resolution images from the frequency values of super-resolution image by the Fourier inverse transformation. The experimental results show that： this reconstruction method can obtain much more clearly super-resolution image in the visual than the recovered super-resolution image through the bilinear interpolation to an arbitrary low- resolution image, almost achieve the same with the actual intake of one times high super-resolution image . PSNR of the restoration of the bridge image can be achieved 34. 767 3 dB. Computer simulations demonstrate the performance of the algorithm. So this method is practicable, and the restored image has good edge visual effects.

关 键 词：傅里叶变换 超分辨恢复 亚像素 频率域 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]