一种亚像素级图像超分辨恢复算法  

A kind of Image Super-Resolution Restoration Algorithm Based on Sub-Pixel

作　　者：赵秀影[1,2] 胡玉臣[2] 翟林培[1] 葛文奇[1] 徐正平[1] 王敬梅 

机构地区：[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]空军航空大学,吉林长春130022 [3]青岛城市建设投资集团有限责任公司,山东青岛266071

出　　处：《光电技术应用》2009年第5期54-56,80,共4页Electro-Optic Technology Application

基　　金：国家863计划(2007AA701203)

摘　　要：随着光学成像到光电数字成像的转变,如何提高CCD的几何分辨率,已成为研制高分辨光电成像系统亟待解决的问题.建立了半像元超分辨成像数学模型,提出了半像元的CCD几何超分辨方法.该方法将2片线阵CCD集成在同一器件中,在线阵方向上错开半个像元,同时读出时间减半,最终交织重组图像数据,以合成高分辨率图像.利用MATLAB软件对双线性插值方法及半像元成像方法进行了仿真,并定性定量地分析了2种方法的效果.结果显示:半像元方法合成图像分辨率约为低分辨率图像的2倍,且2组仿真图像中的PSNR比双线性插值图像分别高出1.486 4 dB和2.207 0 dB.该方法可以显著地减轻欠采样引起的图像模糊,且实时性优于双线性插值方法.With the changes from optical imaging to digital imaging, how to improve the geometric resolution of CCD has become a problem urgent to be solved in developing a high-resolution electrical imaging system. A mathematic model for imaging on super-resolution and subpixel was built, a method to improve the geometric resolution of CCD based on sub-pixel was presented, in which two linear CCDs staggered by half dimension of the pixel in the line direction are integrated in one component, while the reading out time is reduced to half; At last, the image data is interweaved to synthesize a high resolution image. Two methods, bilinear interpolation and sub-pixel imaging, were simulated with the software MATLAB. The effects of the two methods were analyzed qualitatively and quantitatively. The results show that with sub-pixel imaging method, the resolution of synthesized image is twice as large as that of low-resolution image, and the PSNRs in the two groups of simulating images are enhanced by 1. 486 4 dB and 2. 207 0 dB compared to that in bilinear interpolated image. Subpixel imaging method could mitigate greatly the image blur induced by under sampling, and the elapsed time is less than the bilinear interpolation method.

关 键 词：半像元 超分辨 CCD 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]