多光谱与全色遥感图像融合算法的研究
|
摘要
对于大多数遥感应用,为了探测地球表面上的优良特性,人们需要空间分辨率最佳的图像。此外,为了区分不同地面覆盖物,高光谱分辨率图像也同样是最需要的。现代遥感卫星提供的图像,或者是高空间分辨率如全色图像,或者是高光谱图像如多光谱图像。然而,对于某些应用,需要改善多光谱图像的空间分辨率。这个过程叫做锐化。这个过程需要在一个更高的空间分辨率图像上来合成多光谱图像,而这个高空间分辨率图像是由另一种方式获取的高分辨率的图像。图像融合的目的是获取更高的质量信息,但是“更高的质量”的确切定义要取决于实际应用。在遥感领域的图像融合算法,其主要目的就是保持数据的光谱和空间分辨率的保真度。通过比较不同的图像融合技术分析显示,虽然融合图像的质量有了很大的提高,但是上述要求的问题是仍需要进一步关注和研究。
本文主要是以多光谱与全色图像融合为核心领域而展开的研究。主要进行的研究工作有以下几点：
(1)分析多光谱与全色图像融合的研究背景、研究意义以及研究现状,列举一些目前在图像融合领域存在的问题,并介绍今后的发展方向。
(2)介绍图像融合领域最常用的质量评价方法和传统的几种算法,包含了主观和客观的质量评价方法和基于HIS变换的融合算法、基于PCA变换的融合算法、基于lαβ变换的融合算法与基于Brovey变换的融合算法。最后,对这些传统的融合方法进行了实验分析,并依据质量评价标准对这几种融合方法进行了评价,引入了基于HIS变换的融合框架。
(3)对平移不变离散小波变换SIDWT的研究,首先介绍了传统的小波变换,分析了小波变换的优缺点,引出了SIDWT变换。并根据SIDWT变换理论,提出了基于区域特征的SIDWT融合方法,着重对融合规则进行了详述,给出了融合流程,并进行实验分析和对比,验证了该算法的有效性。
(4)详细阐述了拉普拉斯变换和基于梯度的结构相似度GSSIM理论,基于此,提出了基于遗传算法的拉普拉斯图像融合算法。对融合规则进行了详述,并对第一层使用了遗传算法进行融合,对遗传算法中的标函数即适应度函数进行了详述,还有复制、交叉、变异因子等,对其他塔层使用了区域最大能量的方法进行融合；最后通过实验结果与分析,证明了该方法的有效性。
It is desirable to have the best possible spatial resolution in order to detect fine features on the Earth's surface for most remote sensing applications. Moreover, a high spectral resolution is also required to discriminate among different ground covers. The problem with the images provided by modern satellites is that they have either high spatial resolution, i.e., panchromatic image or high spectral resolution, i.e., multi-spectral images. However, for certain applications, there is a need to improve the spatial resolution of the multi-spectral (MS) images. This process is usually called pan-sharpening. This process includes the synthesis of panchromatic (PAN) images at a higher spatial resolution by the help of an alternate high-resolution image obtained by another modality. Image fusion aims at obtaining information of a greater quality, although the exact definition of "greater quality" will depend on the "application". The main obligation of an image fusion algorithm in remote sensing is both the spectral and spatial fidelity of the data. However, recent comparative analyses of different image fusion techniques have shown that, although the quality of fused image has been greatly improved, but the problem is still the need for further attention and study.
The research of this paper is focused on the MS and PAN remote image fusion. The main research works are carried out as the following.
(1)We analyze the research background, significance and research status of the MS and PAN image fusion, and list some of the problems in the field of image fusion, and describe the future direction of development.
(2)We describe some of the most commonly used quality assessment methods of image fusion field and several traditional algorithms about image fusion. These contain the subjective and objective quality evaluation methods, and image fusion algorithm based on HIS transform, and image fusion algorithm based on PCA transform, and image fusion algorithm based on lαβ transform, and image fusion algorithm based Brovey transform. Then, we do some experiments about these traditional image fusion algorithms and analyze the results of the experiments. Last, we evaluate experiments according to the quality evaluation criteria, and introduce a framework based on the integration of HIS transform.
(3) The research of shift-invariant discrete wavelet transform (SIDWT). First, we introduce traditional wavelet transform, and then analyze the advantages and disadvantages of the wavelet transform with SIDWT. Then according to the SIDWT theory, we propose a new image fusion method:SIDWT fusion method based the regional characteristics. We detail the rules fusion, and Draw the flowchart of this new method. Last, to verify the effectiveness of the algorithm, we do some experiments, and analyze it, and compare with others.
(4) We elaborate on the Laplacian transform and gradient-based the structural similarity (GSSIM) theory. First, we introduce the frame of the Laplacian image fusion algorithm and the effectiveness of GSSIM. Second, we propose a new method of the Laplacian image fusion algorithm based on genetic algorithm with GSSIM object function, and detail the fusion rules of this new method, and the first layer using a genetic algorithm. The fitness function (fitness function) of genetic algorithm in the objective function is detailed. Selection, crossover and mutation factor are brief described. To the other tower levels, the area of maximum energy is used. Last, to verify the effectiveness of the algorithm, we do some experiment, and analyze it, and compare with others.
本文主要是以多光谱与全色图像融合为核心领域而展开的研究。主要进行的研究工作有以下几点：
(1)分析多光谱与全色图像融合的研究背景、研究意义以及研究现状,列举一些目前在图像融合领域存在的问题,并介绍今后的发展方向。
(2)介绍图像融合领域最常用的质量评价方法和传统的几种算法,包含了主观和客观的质量评价方法和基于HIS变换的融合算法、基于PCA变换的融合算法、基于lαβ变换的融合算法与基于Brovey变换的融合算法。最后,对这些传统的融合方法进行了实验分析,并依据质量评价标准对这几种融合方法进行了评价,引入了基于HIS变换的融合框架。
(3)对平移不变离散小波变换SIDWT的研究,首先介绍了传统的小波变换,分析了小波变换的优缺点,引出了SIDWT变换。并根据SIDWT变换理论,提出了基于区域特征的SIDWT融合方法,着重对融合规则进行了详述,给出了融合流程,并进行实验分析和对比,验证了该算法的有效性。
(4)详细阐述了拉普拉斯变换和基于梯度的结构相似度GSSIM理论,基于此,提出了基于遗传算法的拉普拉斯图像融合算法。对融合规则进行了详述,并对第一层使用了遗传算法进行融合,对遗传算法中的标函数即适应度函数进行了详述,还有复制、交叉、变异因子等,对其他塔层使用了区域最大能量的方法进行融合；最后通过实验结果与分析,证明了该方法的有效性。
It is desirable to have the best possible spatial resolution in order to detect fine features on the Earth's surface for most remote sensing applications. Moreover, a high spectral resolution is also required to discriminate among different ground covers. The problem with the images provided by modern satellites is that they have either high spatial resolution, i.e., panchromatic image or high spectral resolution, i.e., multi-spectral images. However, for certain applications, there is a need to improve the spatial resolution of the multi-spectral (MS) images. This process is usually called pan-sharpening. This process includes the synthesis of panchromatic (PAN) images at a higher spatial resolution by the help of an alternate high-resolution image obtained by another modality. Image fusion aims at obtaining information of a greater quality, although the exact definition of "greater quality" will depend on the "application". The main obligation of an image fusion algorithm in remote sensing is both the spectral and spatial fidelity of the data. However, recent comparative analyses of different image fusion techniques have shown that, although the quality of fused image has been greatly improved, but the problem is still the need for further attention and study.
The research of this paper is focused on the MS and PAN remote image fusion. The main research works are carried out as the following.
(1)We analyze the research background, significance and research status of the MS and PAN image fusion, and list some of the problems in the field of image fusion, and describe the future direction of development.
(2)We describe some of the most commonly used quality assessment methods of image fusion field and several traditional algorithms about image fusion. These contain the subjective and objective quality evaluation methods, and image fusion algorithm based on HIS transform, and image fusion algorithm based on PCA transform, and image fusion algorithm based on lαβ transform, and image fusion algorithm based Brovey transform. Then, we do some experiments about these traditional image fusion algorithms and analyze the results of the experiments. Last, we evaluate experiments according to the quality evaluation criteria, and introduce a framework based on the integration of HIS transform.
(3) The research of shift-invariant discrete wavelet transform (SIDWT). First, we introduce traditional wavelet transform, and then analyze the advantages and disadvantages of the wavelet transform with SIDWT. Then according to the SIDWT theory, we propose a new image fusion method:SIDWT fusion method based the regional characteristics. We detail the rules fusion, and Draw the flowchart of this new method. Last, to verify the effectiveness of the algorithm, we do some experiments, and analyze it, and compare with others.
(4) We elaborate on the Laplacian transform and gradient-based the structural similarity (GSSIM) theory. First, we introduce the frame of the Laplacian image fusion algorithm and the effectiveness of GSSIM. Second, we propose a new method of the Laplacian image fusion algorithm based on genetic algorithm with GSSIM object function, and detail the fusion rules of this new method, and the first layer using a genetic algorithm. The fitness function (fitness function) of genetic algorithm in the objective function is detailed. Selection, crossover and mutation factor are brief described. To the other tower levels, the area of maximum energy is used. Last, to verify the effectiveness of the algorithm, we do some experiment, and analyze it, and compare with others.
引文
[1]P.K. Varshn. Multi-sensor data fusion. Electronics and Communication Engineering Journal[J].1997,9(6):245-253.
[2]王帅.多源遥感影像数据融合研究[J].科技广场,201 1,(3)：61-63.
[3]陈超,江涛,岳远平等.像素级遥感图像融合方法研究[J].国土资源信息化.2008,(5)：30-34
[4]Wald L. Some terms of reference in data fusion [J].Geo-science and Remote Sensing,IEEE Transactionson.1999,37(3):1190.119.
[5]Daily M.L, Farr T, ElachiC. Geologic Interpretation from Composited Radar and Landsat Imagery[J]. Photogrammetric Engineering and Remote Sensing,1979,45(8):1109-1116.
[6]Yocky D.A. Multi-resolution Wavelet Decomposition Image Merger of Landsat Thematic Mapper and SPOT Panchromatic Data[J]. Photogrammetric Engineering and Remote Sensing.1996.62:1067-1074.
[7]PohlC,Genderen L. Multi-sensor Image Fusion in Remote Sensing:Concepts, Methods and Applications[J]. International Joumal of Remote Sensing,1998,19(5):823-854.
[8]郭雷,李晖晖,包永生.图像融合[M].北京：电子工业出版社,2008：21-27.
[9]易善桢,李琦,杨超伟.数字地球的信息融合框架[J].中国图象图形学报,1999,4A(4)：196-199.
[10]洪日昌.多源图像融合算法及应用研究[D].合肥：中国科技大学,2007.
[11]毛士艺,赵巍.多传感器图像融合技术综述[J].北京航空航天大学学报,2002,28(5)：512-518.
[12]W.J. Carper, T.M. Lillesand, R.W. Kiefer. The use of intensity-hue-saturation transforms for merging SPOT Panchromatic and multispectral image data. Photogrammetric Engineering and Remote Sensing,1990,56:459-467.
[13]Chavez, P.S., Sildes, S.C., Anderson, J. A.,1991. Comparison of three different methods to merge multi-resolution and multi-spectral data:Landsat TM and SPOT panchromatic Photogram. Eng. Rem. Sens.57(3),295-303.
[14]周会珍,吴双,毛德发等.多卫星传感器数据的Brovey融合改进方法[J].遥感学报.2012.16(2)：343-360.
[15]J. Zhou, D. L. Civco, and J. A. Silander, A wavelet transform method to merge Landsat TM and SPOT panchromatic data[J]. Int. J. Remote Sens.,vol.19, no.4, pp.743-757,1998.
[16]P. S. Chavez and A. Y. Kwarteng, Extracting spectral contrast in Landsat Thematic Mapper image data using selective principle component analysis[J] Photogramm. Eng. Remote Sens, vol.55, no.3, pp. 339-348,1989.
[17]Chen M, Xuan J H, Li D R, et al. Image Fusion Algorithm Based on Independent Component Analysis[J]. Opto-Electronic Engineering,2007,34(6):82-87.
[18]Ruderman,D.L., Cronin,T.W., Chiao, C.C., Statistics of cone responses to natural images: Implications for visual coding[J]. J. Opt. Soc. Amer.1998.15 (8),2036-2045.
[19]Reinhard, E., Ashikhmin, M., Gooch, B., Shirley, P.,2001.Color transfer between images[J]. IEEE Computer Graph.Appl.21 (5),34-41.
[20]J.P.Burt, E.H.Andelson. The laplaeian Pyramid as a compact image code[J]. IEEE Trans.on communication,1983,31(4):532-540.
[21]E. H. Adelson, CH. Anderson, J. R. Bergen, P.J. Burt, J. M. Ogden. Pyramid methods in image processing [J]. RCA Engineer, Nov/Dec 1984,29-6.
[22]S.GA.Mallat. Theory for multi-resolution signal decomposition:the wavelet representation [J]. IEEE Trans. Pattern Anal. Machine Intell,1989,11 (7):674-693.
[23]I.Daubechies. The wavelet transform, time-frequency localization and signal analysis. IEEE Trans. Inform Theory,1990,36(5):961-1003.
[24]李鹏,冀晓燕,喻罡等基于平移不变离散小波的多普勒血流信号杂波抑制[J].西安交通大学学报.2007.4(41)：402.406
[25]I.W. Selesnick, R.G. Baraniuk, Kingsbury N G. The dual-tree complex wavelet transform[J]. IEEE Signal Processing Magazine,2005(12):123-151.
[26]Rockinger O. Image Sequence Fusion Using a Shift-invariant Wavelet Transform[C] //Proc. of IEEE International Conference on Image Processing. [S. I.]:IEEE Press,1997.
[27]余连生,文贡坚,李智勇.基于SIDWT的遥感图像融合算法[J].计算机工程.2011(17)：1 97-1 99
[28]Q.S.Sun, J.Zhong, P.A.Heng,et.A novel feature fusion method based on Partial Least Squares Regression [J]. Proceedings ofICAPR05.2003.268-277.
[29]GBebis, A.Gyaourova, S.Singh,et. Face recognition by fusing thermal infrared and visible Imagery[J].Image and Vision Computing,2006,24(7):727-742.
[30]胡良梅,高隽,何柯峰.图像融合质量评价方法的研究[J].电子学报.2004.12(24)：218-221.
[31]敬忠良,肖刚,李振华.图像融合-理论与应用[M].北京：高等教育出版社,2007.
[32]王海晖,彭嘉雄,吴巍等.多源遥感图像融合效果评价方法研究[J].计算机工程与应用.2003.25：33-37.
[33]李树涛,王耀南,张昌凡.多传感器图像融合的客观评价与分析[J].仪表学报.2002,23(6)：651-654.
[34]习夏明革,何友,黄晓东.多传感器图像融合效果评价方法研究[J].电光制.2003,10(2)：31-35.
[35]霍宏涛.数字图像处理[M].北京：机械工业出版社,2003.184-209.
[36]伍娟,卢凌,董静.基于HIS变换与直方图匹配法的遥感图像融合[J].武汉理工大学学报.2004.2(28)：55-58
[37]Tian J,Chen J,Zhang C. Multispectral image fusion based on fractal feature[C]//Presceedings of SPIE. Bellingham:Socety of Photo-Optical Instrumentation Engineers Press,2004,5308:824-832
[38]罗希平,田捷,诸葛婴等.图像分割方法综述[J].模式识别与人工智能,1999,9(12)：300-312
[39]苗启广,王宝树.基于改进的拉普拉斯金字塔变换的图像融合方法[J].光学学报.2007,9(27)：1605.1610
[40]F. W. Campbell,J.G. Rob son. Application of Fourier analysis to the visibility of gratings[J]. J.Physiology,1968,197 (3):55-556
[41]Zhou W, Bovik A. C, Li gang L. Why is image quality assessment so difficult? [C]// Proc of IEEE International Conference on Acoustics,Speech,and Signal Processing (IC AS SP'02),2002. Orlando, FL,USA,2002:IV 3313.3316.
[42]Zhou W, Bovik A.C., Sheikh H R,et al. Image quality assessment:from error visibility to structural similarity[J].IEEE Trans Image Process,2004,13(4):600.612.
[43]Chen Guanhao, Yang Chunling,Member IEEE. Gradient-based structural similarity for image Quality Assessment[J].IEEE,ICIP.2006,2929-2932.
[44]彭开.基于遗传算法的遥感图像融合方法研究[D].西安：电子科技大学,2010.
[45]段玉倩,贺家李.遗传算法及其改进[J].电力系统及其自动化学报.1998,3(10)：39-52.
[46]Schaffer J.D. A Study of Control Parameters Affecting Online Performance of Genetic Algorithms for Function Optimization. Proc.3rd.Conf.Genetic Algorithms.1989:51-60.
[47]杨龙.红外热像和可见光图像融合[D].合肥：安徽大学,2012.
[2]王帅.多源遥感影像数据融合研究[J].科技广场,201 1,(3)：61-63.
[3]陈超,江涛,岳远平等.像素级遥感图像融合方法研究[J].国土资源信息化.2008,(5)：30-34
[4]Wald L. Some terms of reference in data fusion [J].Geo-science and Remote Sensing,IEEE Transactionson.1999,37(3):1190.119.
[5]Daily M.L, Farr T, ElachiC. Geologic Interpretation from Composited Radar and Landsat Imagery[J]. Photogrammetric Engineering and Remote Sensing,1979,45(8):1109-1116.
[6]Yocky D.A. Multi-resolution Wavelet Decomposition Image Merger of Landsat Thematic Mapper and SPOT Panchromatic Data[J]. Photogrammetric Engineering and Remote Sensing.1996.62:1067-1074.
[7]PohlC,Genderen L. Multi-sensor Image Fusion in Remote Sensing:Concepts, Methods and Applications[J]. International Joumal of Remote Sensing,1998,19(5):823-854.
[8]郭雷,李晖晖,包永生.图像融合[M].北京：电子工业出版社,2008：21-27.
[9]易善桢,李琦,杨超伟.数字地球的信息融合框架[J].中国图象图形学报,1999,4A(4)：196-199.
[10]洪日昌.多源图像融合算法及应用研究[D].合肥：中国科技大学,2007.
[11]毛士艺,赵巍.多传感器图像融合技术综述[J].北京航空航天大学学报,2002,28(5)：512-518.
[12]W.J. Carper, T.M. Lillesand, R.W. Kiefer. The use of intensity-hue-saturation transforms for merging SPOT Panchromatic and multispectral image data. Photogrammetric Engineering and Remote Sensing,1990,56:459-467.
[13]Chavez, P.S., Sildes, S.C., Anderson, J. A.,1991. Comparison of three different methods to merge multi-resolution and multi-spectral data:Landsat TM and SPOT panchromatic Photogram. Eng. Rem. Sens.57(3),295-303.
[14]周会珍,吴双,毛德发等.多卫星传感器数据的Brovey融合改进方法[J].遥感学报.2012.16(2)：343-360.
[15]J. Zhou, D. L. Civco, and J. A. Silander, A wavelet transform method to merge Landsat TM and SPOT panchromatic data[J]. Int. J. Remote Sens.,vol.19, no.4, pp.743-757,1998.
[16]P. S. Chavez and A. Y. Kwarteng, Extracting spectral contrast in Landsat Thematic Mapper image data using selective principle component analysis[J] Photogramm. Eng. Remote Sens, vol.55, no.3, pp. 339-348,1989.
[17]Chen M, Xuan J H, Li D R, et al. Image Fusion Algorithm Based on Independent Component Analysis[J]. Opto-Electronic Engineering,2007,34(6):82-87.
[18]Ruderman,D.L., Cronin,T.W., Chiao, C.C., Statistics of cone responses to natural images: Implications for visual coding[J]. J. Opt. Soc. Amer.1998.15 (8),2036-2045.
[19]Reinhard, E., Ashikhmin, M., Gooch, B., Shirley, P.,2001.Color transfer between images[J]. IEEE Computer Graph.Appl.21 (5),34-41.
[20]J.P.Burt, E.H.Andelson. The laplaeian Pyramid as a compact image code[J]. IEEE Trans.on communication,1983,31(4):532-540.
[21]E. H. Adelson, CH. Anderson, J. R. Bergen, P.J. Burt, J. M. Ogden. Pyramid methods in image processing [J]. RCA Engineer, Nov/Dec 1984,29-6.
[22]S.GA.Mallat. Theory for multi-resolution signal decomposition:the wavelet representation [J]. IEEE Trans. Pattern Anal. Machine Intell,1989,11 (7):674-693.
[23]I.Daubechies. The wavelet transform, time-frequency localization and signal analysis. IEEE Trans. Inform Theory,1990,36(5):961-1003.
[24]李鹏,冀晓燕,喻罡等基于平移不变离散小波的多普勒血流信号杂波抑制[J].西安交通大学学报.2007.4(41)：402.406
[25]I.W. Selesnick, R.G. Baraniuk, Kingsbury N G. The dual-tree complex wavelet transform[J]. IEEE Signal Processing Magazine,2005(12):123-151.
[26]Rockinger O. Image Sequence Fusion Using a Shift-invariant Wavelet Transform[C] //Proc. of IEEE International Conference on Image Processing. [S. I.]:IEEE Press,1997.
[27]余连生,文贡坚,李智勇.基于SIDWT的遥感图像融合算法[J].计算机工程.2011(17)：1 97-1 99
[28]Q.S.Sun, J.Zhong, P.A.Heng,et.A novel feature fusion method based on Partial Least Squares Regression [J]. Proceedings ofICAPR05.2003.268-277.
[29]GBebis, A.Gyaourova, S.Singh,et. Face recognition by fusing thermal infrared and visible Imagery[J].Image and Vision Computing,2006,24(7):727-742.
[30]胡良梅,高隽,何柯峰.图像融合质量评价方法的研究[J].电子学报.2004.12(24)：218-221.
[31]敬忠良,肖刚,李振华.图像融合-理论与应用[M].北京：高等教育出版社,2007.
[32]王海晖,彭嘉雄,吴巍等.多源遥感图像融合效果评价方法研究[J].计算机工程与应用.2003.25：33-37.
[33]李树涛,王耀南,张昌凡.多传感器图像融合的客观评价与分析[J].仪表学报.2002,23(6)：651-654.
[34]习夏明革,何友,黄晓东.多传感器图像融合效果评价方法研究[J].电光制.2003,10(2)：31-35.
[35]霍宏涛.数字图像处理[M].北京：机械工业出版社,2003.184-209.
[36]伍娟,卢凌,董静.基于HIS变换与直方图匹配法的遥感图像融合[J].武汉理工大学学报.2004.2(28)：55-58
[37]Tian J,Chen J,Zhang C. Multispectral image fusion based on fractal feature[C]//Presceedings of SPIE. Bellingham:Socety of Photo-Optical Instrumentation Engineers Press,2004,5308:824-832
[38]罗希平,田捷,诸葛婴等.图像分割方法综述[J].模式识别与人工智能,1999,9(12)：300-312
[39]苗启广,王宝树.基于改进的拉普拉斯金字塔变换的图像融合方法[J].光学学报.2007,9(27)：1605.1610
[40]F. W. Campbell,J.G. Rob son. Application of Fourier analysis to the visibility of gratings[J]. J.Physiology,1968,197 (3):55-556
[41]Zhou W, Bovik A. C, Li gang L. Why is image quality assessment so difficult? [C]// Proc of IEEE International Conference on Acoustics,Speech,and Signal Processing (IC AS SP'02),2002. Orlando, FL,USA,2002:IV 3313.3316.
[42]Zhou W, Bovik A.C., Sheikh H R,et al. Image quality assessment:from error visibility to structural similarity[J].IEEE Trans Image Process,2004,13(4):600.612.
[43]Chen Guanhao, Yang Chunling,Member IEEE. Gradient-based structural similarity for image Quality Assessment[J].IEEE,ICIP.2006,2929-2932.
[44]彭开.基于遗传算法的遥感图像融合方法研究[D].西安：电子科技大学,2010.
[45]段玉倩,贺家李.遗传算法及其改进[J].电力系统及其自动化学报.1998,3(10)：39-52.
[46]Schaffer J.D. A Study of Control Parameters Affecting Online Performance of Genetic Algorithms for Function Optimization. Proc.3rd.Conf.Genetic Algorithms.1989:51-60.
[47]杨龙.红外热像和可见光图像融合[D].合肥：安徽大学,2012.