亚像元的CCD几何超分辨方法  

CCD geometric superresolution method based on subpixel

作　　者：徐正平[1,2] 翟林培[1] 葛文奇[1] 赵秀影[1,3] 刘妍妍[1,2] 耿文豹[1,2] 

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]中国科学院研究生院,北京100039 [3]空军航空大学,吉林长春130022

出　　处：《光学精密工程》2008年第12期2447-2453,共7页Optics and Precision Engineering

基　　金：中国科学院长春光学精密机械与物理研究所创新工程资助项目(No.ZJ99130B)

摘　　要：为了在现有CCD像元尺寸的基础上提高CCD像元分辨力,对CCD几何超分辨问题进行了研究。从研究现状入手,给出了现有算法及其不足,建立了亚像元超分辨成像数学模型,提出了亚像元的CCD几何超分辨方法。该方法将两片线阵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MATLAB7.0.1软件对双线性插值方法及亚像元成像方法进行了仿真,并定性定量地分析了两种方法的效果。结果显示:亚像元方法合成图像分辨率约为低分辨率图像的2倍,且两组仿真图像中的PSNR比双线性插值图像分别高出1.4864dB和2.2070dB。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。In order to enhance the resolution of CCD pixel while keeping pixel dimension, the problem of CCD geometric superresolution is researched. Beginning with the actuality of superresolution, the existing methods and their shortcomings are presented, and a mathematical model of the subpixel superresolution is established. A method to improve the geometric resolution of CCD based on subpixel is presented,in which two linear CCDs staggered by half dimension of the pixel in the line direction are integrated in one component. When operated, 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, are simulated with the software MATLAB7. 0.1. The qualitative and quantitative analysis are given. The simulation 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 with that in bilinear interpolated image. Sub-pixel 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：电荷耦合器件 亚像元 几何超分辨 

分 类 号：TN386.5[电子电信—物理电子学]