基于序列图像重构的超分辨率成像技术研究
|
摘要
基于序列图像的超分辨率图像重构就是由一序列互有位移的低分辨率图像来估计一幅(或一序列)较高分辨率的图像，同时还能够消除加性噪声以及由有限检测器尺寸和光学元件产生的模糊。由于该方法既克服了通过减小增加感光单元尺寸来提高CCD分辨率所面对的工艺难题，又能够利用多幅图像的不同信息达到获取更高分辨率图像的目的，成为图像复原领域中的一个重要分支。目前，这项技术广泛应用于遥感、医学成像和高清晰度电视等多个领域。基于序列图像重构的超分辨率成像技术研究主要集中在两个方面：①空间上互有位移的序列初始图像的获取；②高分辨率图像的重构算法。
论文首先介绍了课题的背景，超分辨率图像重构在国内外的发展状况及本课题要解决的问题和达到的要求。然后分析了各种超分辨率复原算法，最后得出了较简单易行的，能从真正意义上提高分辨率的算法—反演解析法，引出课题。接着设计两种图像获取方案，并对其原理和结构进行了详细的分析；再下来便是主要的实验部分，在对图像接收器件的分析和选择后，分别做了一维方向提高1.5倍、2倍和2.5倍的实验，得到了这样的结论：对于帧转移CCD，我们使用反演解析法，利用两幅有2/3象素位移的原图重构分辨率提高1.5倍的高分辨率图，实验效果还是比较理想的。对于行转移CCD，针对其盲区较大的特点，可以设计较简单的象素交叉插值，利用两幅位移为0.5象素的原图，获得分辨率提高2倍的高分辨率图。而由于CCD像元尺寸与弥散斑相差不多，要想获得分辨率提高2.5倍的高分辨率图已经很困难，说明反演解析法实际上对分辨率的提高是有一个上限的，这和弥散斑与CCD感光像元间的大小比例有关。接着针对图像接收器—CCD的噪声源类型分析了不同噪声对实验结果的影响，并用软件的方法成功地对噪声进行了平衡。另外还分析了噪声对分辨率提高倍数的影响，通过仿真实验，找到了合成图像受随机高斯白噪声影响的规律。最后对本课题进行了总结与展望，提出了系统的改进方向。
Super-resolution reconstruction basing on the image sequence is a technique for estimating a high-resolution image (or sequence) from an image sequence having slight displacement and combating additive noise and blurring due to the finite detector size and optics. As this technique not only conquers the difficulty of enhancing the CCD resolution from reducing the size of CCD sensor' cell,but also gets the high-resolution image using the different information of multi images,it has become the important embranchment of image restoration. At present,this technique is widely used for remote-sensing,medical imaging and high-definition television (HDTV) standard,etc. Now,the research of super-resolution reconstruction technique basing on the image sequence focuses on two factors:(D capturing an image sequence with slight displacement;﹖he reconstruction algorithm of high-resolution image.
The thesis introduces the background,development status and object of our research firstly. The second step,many kinds of super-resolution reconstruction algorithms are discussed. We find the method of inversion analysis is a more effective and convenient one in these algorisms. The third step,we propose two methods for image capturing. Then the principle and machinery of them are discussed. The fourth step,the experiments of the enhancement ratio of 1.5,2 and 2.5 are carried out. The method of inversion analysis is used for FT CCD and the method of simple pixel cross-interpolation is used for IT CCD. From the results of these experiments,we know that the improvement effect is clear when the enhancement ratio is 1.5. But the enhancement ratio of 2.5 is difficult to be reached because of the optical dispersion. It shows that the enhancement ratio is limited by the size of CCD sensor's cell and optical dispersion. The fifth step,we analyze the influence on the experiments result by a few different kinds of noi
se and propose an algorism to decrease it. In addition,the influence mechanism of noise on the enhancement ratio is discussed. From the simulation examples,we find the law that random gaussian white noise influencing synthesized images. Finally,a summary for the thesis is given and the direction of improvement is
proposed.
论文首先介绍了课题的背景，超分辨率图像重构在国内外的发展状况及本课题要解决的问题和达到的要求。然后分析了各种超分辨率复原算法，最后得出了较简单易行的，能从真正意义上提高分辨率的算法—反演解析法，引出课题。接着设计两种图像获取方案，并对其原理和结构进行了详细的分析；再下来便是主要的实验部分，在对图像接收器件的分析和选择后，分别做了一维方向提高1.5倍、2倍和2.5倍的实验，得到了这样的结论：对于帧转移CCD，我们使用反演解析法，利用两幅有2/3象素位移的原图重构分辨率提高1.5倍的高分辨率图，实验效果还是比较理想的。对于行转移CCD，针对其盲区较大的特点，可以设计较简单的象素交叉插值，利用两幅位移为0.5象素的原图，获得分辨率提高2倍的高分辨率图。而由于CCD像元尺寸与弥散斑相差不多，要想获得分辨率提高2.5倍的高分辨率图已经很困难，说明反演解析法实际上对分辨率的提高是有一个上限的，这和弥散斑与CCD感光像元间的大小比例有关。接着针对图像接收器—CCD的噪声源类型分析了不同噪声对实验结果的影响，并用软件的方法成功地对噪声进行了平衡。另外还分析了噪声对分辨率提高倍数的影响，通过仿真实验，找到了合成图像受随机高斯白噪声影响的规律。最后对本课题进行了总结与展望，提出了系统的改进方向。
Super-resolution reconstruction basing on the image sequence is a technique for estimating a high-resolution image (or sequence) from an image sequence having slight displacement and combating additive noise and blurring due to the finite detector size and optics. As this technique not only conquers the difficulty of enhancing the CCD resolution from reducing the size of CCD sensor' cell,but also gets the high-resolution image using the different information of multi images,it has become the important embranchment of image restoration. At present,this technique is widely used for remote-sensing,medical imaging and high-definition television (HDTV) standard,etc. Now,the research of super-resolution reconstruction technique basing on the image sequence focuses on two factors:(D capturing an image sequence with slight displacement;﹖he reconstruction algorithm of high-resolution image.
The thesis introduces the background,development status and object of our research firstly. The second step,many kinds of super-resolution reconstruction algorithms are discussed. We find the method of inversion analysis is a more effective and convenient one in these algorisms. The third step,we propose two methods for image capturing. Then the principle and machinery of them are discussed. The fourth step,the experiments of the enhancement ratio of 1.5,2 and 2.5 are carried out. The method of inversion analysis is used for FT CCD and the method of simple pixel cross-interpolation is used for IT CCD. From the results of these experiments,we know that the improvement effect is clear when the enhancement ratio is 1.5. But the enhancement ratio of 2.5 is difficult to be reached because of the optical dispersion. It shows that the enhancement ratio is limited by the size of CCD sensor's cell and optical dispersion. The fifth step,we analyze the influence on the experiments result by a few different kinds of noi
se and propose an algorism to decrease it. In addition,the influence mechanism of noise on the enhancement ratio is discussed. From the simulation examples,we find the law that random gaussian white noise influencing synthesized images. Finally,a summary for the thesis is given and the direction of improvement is
proposed.
引文
1. Aizawa, K. Komatsu, T. and Saito, T. A Scheme for Acquiring Very High Resolution Images Using Multiple Cameras. IEEE International Conference on Acoustics, Speech, and Signal Processing, 1992,3:289-292
2．陈春晓．压电陶瓷体(PZT)制作及PZT微位移装置．光仪技术，1999，15(2)：31-35
3. Fryer, J., and McIntosh, K.. Enhancement of Image Resolution in Digital Photogrammtry. Photogrammetric Engineering and Remote Sensing, 2001, 67(6): 741-749
4. Godding, R., and D. Woytowicz. A New Digital High Resolution Recording System. International Archives of Photogrammetry and Remote Sensing, 1995,30(5): 31-35
5．黄战华，蔡怀宇．基于半象素错位的多幅图像重建高分辨率图像技术研究．光学技术，2002，28(5)：389-391，394
6. Hunt B R. Super-Resolution of Image: Algorithms, Principles, Performance[J]. International Journal of Imaging Systems and Technology, 1995, 6:297-304
7．金杰，徐锡林．提高CCD分辨率的一种尝试．现代计量测试，1997，5(3)：37-41
8．刘良云，李英才．超分辨率图像重构技术的仿真实验研究．中国图像图形学报，2001，6(7)：629-635
9．刘新平，高瞻．面阵CCD作探测器的“亚像元”成像方法及实验．科学通报，1999，44(15)：1063-1065
10．麦伟麟．光学传递函数极其数理基础．北京：国防工业出版社，1979
11. Michael K. Ng, N. K. Bose. Analysis of Displacement Errors in High-Resolution Image Reconstruction With Multisensors. IEEE Transactions on Circuits and Systems, 2002, 49(6): 806-813
12. Numnonda, T. Andrews, M.. High Resolution Image Reconstruction Using Mean Field Annealing. Neural Networks for Signal Processing [1994] IV. Proceedings of the 1994 IEEE Workshop, 1994, 6(8): 441-450
13. Patti, A., M.I. Sezan. And A.M. Tekaio. Superresolution video reconstruction with arbitrary sampling lattices and nonzero aperture time. IEEE Transactions on Image Processing, 1997,6(8): 1064-1076
14. Saito, T., Komatsu, T., Aizawa, K.. Very High Resolution Image Acquisition Through Image Reconstruction From Lower Resolution Images Taken With Multiple Cameras. Singapore ICCS/ISITA'92. 'Communications on the Move', 1990,2:639-643
15．邵凌，丁佩律．重建高分辨率图像的实时串行迭代算法．红外与毫米波学报，2002，21(2)：104-108
16. Simon Baker, Takeo Kanade. Limist on Super-Resolution and How to Break Them. IEEE Transaction on Pattern Analysis and Machine Intelligence, 2002, 24(9): 1167-1183
17．苏秉华，金伟其．超分辨率图像复原极其进展．光学技术，2001，27(1)：6-9
18．苏秉华，金伟其，等．基于MarKov约束的泊松最大后验概率超分辨率图像复原法．光子学报，2002，31(4)：492-496
19．谈新权．CCD高分辨率成像技术(一)．电子与自动化，1999，28(2)：52-54
20．谈新权．CCD高分辨率成像技术(二)．电子与自动化，1999，28(3)：53-55
21．谈新权．CCD高分辨率成像技术(三)．电子与自动化，1999，28(4)：52-55
22．谈新权．CCD高分辨率成像技术(四)．电子与自动化，1999，28(5)：52-55
23．谈新权．CCD空间偏置成像技术．激光与红外，1998，28(3)：169-172
24. Tekalp A M, Ozkan M K, Sezan M I. High-Resolution Image Reconstruction form Lower-Resolution Image Sequence and Space-Varying Image Restoratong[J]. J.O.S.A., 1989,6(11): 1715-1726
25. Tsai R Y, Huang T S. Multiframe image restoration and restriction[J]. Advances in Computer Vision and Image Processing. 1984,1: 317～319
26．王凌，张平．基于多次移位成像提高CCD成像分辨率的反演解析法．光电工程(已录用)
27．王凌，张平．多次移位成像降低噪声对反演解析法的影响研究．光电子激光(已录用)
28．王凌，张平．反演解析法提高CCD成像分辨率的噪声影响研究．计量学报(已录用)
29．王凌，张平．Solution for Negative Influence of Exposure Dissimilarity on High-Resolution Image Reconstruction．中国光学快报(已录用)
30．王之江，伍树东．成像光学．北京：科学出版社，1991
31．徐之海，李奇．现代成像系统[M]．北京：国防工业出版社，2001．
32．张利，王彦．一种用四摄像机合拍高分辨率图像的方法．电视技术，1998，(12)：18-20
33．张平，王凌．光纤连接器端面检测技术．光学仪器，2002，24(3)：3-7
34．张平，王凌．基于多次成像的分辨率提高算法研究．光子学报(已录用)
35．张新明，沈兰荪．超分辨率复原技术的发展．测控技术，2002，21(5)：33-35
36．赵辉，浦昭邦．压电器件在微位移系统中的应用．压电与声光，2000，22(4)：244-245，252
37．庄松林，钱振邦．光学传递函数．北京：机械工业出版社，1981
2．陈春晓．压电陶瓷体(PZT)制作及PZT微位移装置．光仪技术，1999，15(2)：31-35
3. Fryer, J., and McIntosh, K.. Enhancement of Image Resolution in Digital Photogrammtry. Photogrammetric Engineering and Remote Sensing, 2001, 67(6): 741-749
4. Godding, R., and D. Woytowicz. A New Digital High Resolution Recording System. International Archives of Photogrammetry and Remote Sensing, 1995,30(5): 31-35
5．黄战华，蔡怀宇．基于半象素错位的多幅图像重建高分辨率图像技术研究．光学技术，2002，28(5)：389-391，394
6. Hunt B R. Super-Resolution of Image: Algorithms, Principles, Performance[J]. International Journal of Imaging Systems and Technology, 1995, 6:297-304
7．金杰，徐锡林．提高CCD分辨率的一种尝试．现代计量测试，1997，5(3)：37-41
8．刘良云，李英才．超分辨率图像重构技术的仿真实验研究．中国图像图形学报，2001，6(7)：629-635
9．刘新平，高瞻．面阵CCD作探测器的“亚像元”成像方法及实验．科学通报，1999，44(15)：1063-1065
10．麦伟麟．光学传递函数极其数理基础．北京：国防工业出版社，1979
11. Michael K. Ng, N. K. Bose. Analysis of Displacement Errors in High-Resolution Image Reconstruction With Multisensors. IEEE Transactions on Circuits and Systems, 2002, 49(6): 806-813
12. Numnonda, T. Andrews, M.. High Resolution Image Reconstruction Using Mean Field Annealing. Neural Networks for Signal Processing [1994] IV. Proceedings of the 1994 IEEE Workshop, 1994, 6(8): 441-450
13. Patti, A., M.I. Sezan. And A.M. Tekaio. Superresolution video reconstruction with arbitrary sampling lattices and nonzero aperture time. IEEE Transactions on Image Processing, 1997,6(8): 1064-1076
14. Saito, T., Komatsu, T., Aizawa, K.. Very High Resolution Image Acquisition Through Image Reconstruction From Lower Resolution Images Taken With Multiple Cameras. Singapore ICCS/ISITA'92. 'Communications on the Move', 1990,2:639-643
15．邵凌，丁佩律．重建高分辨率图像的实时串行迭代算法．红外与毫米波学报，2002，21(2)：104-108
16. Simon Baker, Takeo Kanade. Limist on Super-Resolution and How to Break Them. IEEE Transaction on Pattern Analysis and Machine Intelligence, 2002, 24(9): 1167-1183
17．苏秉华，金伟其．超分辨率图像复原极其进展．光学技术，2001，27(1)：6-9
18．苏秉华，金伟其，等．基于MarKov约束的泊松最大后验概率超分辨率图像复原法．光子学报，2002，31(4)：492-496
19．谈新权．CCD高分辨率成像技术(一)．电子与自动化，1999，28(2)：52-54
20．谈新权．CCD高分辨率成像技术(二)．电子与自动化，1999，28(3)：53-55
21．谈新权．CCD高分辨率成像技术(三)．电子与自动化，1999，28(4)：52-55
22．谈新权．CCD高分辨率成像技术(四)．电子与自动化，1999，28(5)：52-55
23．谈新权．CCD空间偏置成像技术．激光与红外，1998，28(3)：169-172
24. Tekalp A M, Ozkan M K, Sezan M I. High-Resolution Image Reconstruction form Lower-Resolution Image Sequence and Space-Varying Image Restoratong[J]. J.O.S.A., 1989,6(11): 1715-1726
25. Tsai R Y, Huang T S. Multiframe image restoration and restriction[J]. Advances in Computer Vision and Image Processing. 1984,1: 317～319
26．王凌，张平．基于多次移位成像提高CCD成像分辨率的反演解析法．光电工程(已录用)
27．王凌，张平．多次移位成像降低噪声对反演解析法的影响研究．光电子激光(已录用)
28．王凌，张平．反演解析法提高CCD成像分辨率的噪声影响研究．计量学报(已录用)
29．王凌，张平．Solution for Negative Influence of Exposure Dissimilarity on High-Resolution Image Reconstruction．中国光学快报(已录用)
30．王之江，伍树东．成像光学．北京：科学出版社，1991
31．徐之海，李奇．现代成像系统[M]．北京：国防工业出版社，2001．
32．张利，王彦．一种用四摄像机合拍高分辨率图像的方法．电视技术，1998，(12)：18-20
33．张平，王凌．光纤连接器端面检测技术．光学仪器，2002，24(3)：3-7
34．张平，王凌．基于多次成像的分辨率提高算法研究．光子学报(已录用)
35．张新明，沈兰荪．超分辨率复原技术的发展．测控技术，2002，21(5)：33-35
36．赵辉，浦昭邦．压电器件在微位移系统中的应用．压电与声光，2000，22(4)：244-245，252
37．庄松林，钱振邦．光学传递函数．北京：机械工业出版社，1981