基于自适应遗传分割算法的雾天图像处理方法研究
|
摘要
在雾天条件下,户外环境的视觉系统获取的图像的对比度和颜色会出现严重的退化,使得民用领域或是军事领域中很多重要的监控系统难以正常工作。因此在计算机视觉系统中,有必要引进能使雾天图像得到有效处理的机制,对受天气影响的图像进行复原,进而减小天气对图像质量的影响。为了实现视觉系统能在恶劣天气继续全天候工作,提高系统的鲁棒性和可靠性,对雾天条件下获取的退化图像进行复原方法的研究具有现实意义。
论文应用自适应遗传算法对雾天图像进行优化阈值分割,并在此基础上对退化图像进行有效的还原处理。主要工作包括:
第一,将遗传算法的优化特性应用到最佳熵算法、最大类间方差法及Fisher准则函数分割算法中,对基于最佳熵算法的遗传分割算法、基于最大类间方差的遗传分割方法、基于Fisher准则函数算法的三种图像分割方法进行实验仿真比较,验证了基于Fisher准则函数算法的遗传分割方法的实时优越性。
第二,从减少遗传算法陷入局部最优点及提高遗传算法全局搜索能力的角度出发,进一步研究了基于Fisher准则的自适应遗传分割算法。通过实验仿真,得出了分割算法中的最佳遗传算子,并运用该方法对雾天图像进行阈值分割,得到优化的灰度阈值结果。
第三,采用自适应遗传分割算法对雾天图像优化分割成两部分并分别进行图像的复原。由于雾天图像远景部分灰度分布比较集中,接近天空亮度;而近景灰度分布相对均衡,能比较清晰地分辨,因此分别采用McCartney模型及直方图均衡化的恢复方法进行综合处理,得到图像复原结果。通过实验分析和验证,该方法能有效地改善雾天图像的退化现象和提高图像清晰度。此外,运用本文中的方法对各种天气下所拍摄的图像进行仿真实验并对结果进行对比,验证了基于McCartney模型的恢复方法对远景处理的实用性。
最后,对本文的研究工作进行了总结,并提出了有待进一步研究的关键技术和发展方向。
In fog conditions, the image contrast and color which obtained by the outdoor video surveillance systems will have a serious degradation, it makes many important monitoring systems in the civilian or military areas can not work correctly. So in the computer vision systems, it is necessary to introduce the work mechanism which can not only effectively process the image, but also can recover the image to reduce the impact of weather on image quality. In order to make video surveillance systems always operate normally in bad weather, it need improve the robustness and reliability of outdoor surveillance systems, so it is significant to research the image restoration method in the fog condition.
This paper uses the adaptive GA to segment the fog image, and processes the segregated results so the fog image is recovered effectively. The research of this paper includes the works:
1. Because of the optimization features, this paper adopts the genetic algorithm to the best entropy algorithm, Otsu method and the Fisher criterion function segmentation algorithm, and proposes the best-entropy-based genetic segmentation algorithm, the Ostu-based genetic segmentation algorithm and the genetic segmentation method based on Fisher criterion function segmentation algorithm. Through the experiments comparison, the real-time advantage of the genetic segmentation method based on Fisher criterion function segmentation algorithm is superior to the other methods.
2. In order to reduce the chances falling into the local minima and enhance the global search ability, this paper proposes the adaptive GA based on the Fisher criterion function. Through the simulation experiments, it can obtain the best genetic operators, and by using the method which proposed in this paper to segment the fog image, the segregated results of fog is accurate and the image is divided into two parts.
3. As fog have such characteristics: the establishing shot is a gray-scale distribution of some more concentrated and closer to the sky brightness of the region, while the close shot can tell in misty image and the gray-scale distribution is relatively balanced distribution and more clearly. For those two parts, the histogram equalization and the recovery approach based on the McCartney model is used separately. After the comprehensive treatment, the image is recovered well. The experiment shows that the method proposed above is very effective to improve the degradation of images and can enhance the clearness of images. Moreover, using the method to process the image taken in various weather, through the simulation experiments and comparing the results ,it can verified the practicability of the recovery method based on McCartney model to process the establishing shot.
Finally, through summarizing the work, the further study is proposed to be the key technology and development.
论文应用自适应遗传算法对雾天图像进行优化阈值分割,并在此基础上对退化图像进行有效的还原处理。主要工作包括:
第一,将遗传算法的优化特性应用到最佳熵算法、最大类间方差法及Fisher准则函数分割算法中,对基于最佳熵算法的遗传分割算法、基于最大类间方差的遗传分割方法、基于Fisher准则函数算法的三种图像分割方法进行实验仿真比较,验证了基于Fisher准则函数算法的遗传分割方法的实时优越性。
第二,从减少遗传算法陷入局部最优点及提高遗传算法全局搜索能力的角度出发,进一步研究了基于Fisher准则的自适应遗传分割算法。通过实验仿真,得出了分割算法中的最佳遗传算子,并运用该方法对雾天图像进行阈值分割,得到优化的灰度阈值结果。
第三,采用自适应遗传分割算法对雾天图像优化分割成两部分并分别进行图像的复原。由于雾天图像远景部分灰度分布比较集中,接近天空亮度;而近景灰度分布相对均衡,能比较清晰地分辨,因此分别采用McCartney模型及直方图均衡化的恢复方法进行综合处理,得到图像复原结果。通过实验分析和验证,该方法能有效地改善雾天图像的退化现象和提高图像清晰度。此外,运用本文中的方法对各种天气下所拍摄的图像进行仿真实验并对结果进行对比,验证了基于McCartney模型的恢复方法对远景处理的实用性。
最后,对本文的研究工作进行了总结,并提出了有待进一步研究的关键技术和发展方向。
In fog conditions, the image contrast and color which obtained by the outdoor video surveillance systems will have a serious degradation, it makes many important monitoring systems in the civilian or military areas can not work correctly. So in the computer vision systems, it is necessary to introduce the work mechanism which can not only effectively process the image, but also can recover the image to reduce the impact of weather on image quality. In order to make video surveillance systems always operate normally in bad weather, it need improve the robustness and reliability of outdoor surveillance systems, so it is significant to research the image restoration method in the fog condition.
This paper uses the adaptive GA to segment the fog image, and processes the segregated results so the fog image is recovered effectively. The research of this paper includes the works:
1. Because of the optimization features, this paper adopts the genetic algorithm to the best entropy algorithm, Otsu method and the Fisher criterion function segmentation algorithm, and proposes the best-entropy-based genetic segmentation algorithm, the Ostu-based genetic segmentation algorithm and the genetic segmentation method based on Fisher criterion function segmentation algorithm. Through the experiments comparison, the real-time advantage of the genetic segmentation method based on Fisher criterion function segmentation algorithm is superior to the other methods.
2. In order to reduce the chances falling into the local minima and enhance the global search ability, this paper proposes the adaptive GA based on the Fisher criterion function. Through the simulation experiments, it can obtain the best genetic operators, and by using the method which proposed in this paper to segment the fog image, the segregated results of fog is accurate and the image is divided into two parts.
3. As fog have such characteristics: the establishing shot is a gray-scale distribution of some more concentrated and closer to the sky brightness of the region, while the close shot can tell in misty image and the gray-scale distribution is relatively balanced distribution and more clearly. For those two parts, the histogram equalization and the recovery approach based on the McCartney model is used separately. After the comprehensive treatment, the image is recovered well. The experiment shows that the method proposed above is very effective to improve the degradation of images and can enhance the clearness of images. Moreover, using the method to process the image taken in various weather, through the simulation experiments and comparing the results ,it can verified the practicability of the recovery method based on McCartney model to process the establishing shot.
Finally, through summarizing the work, the further study is proposed to be the key technology and development.
引文
[1]杨枝灵,王开.数字图像获取处理及实践应用.北京:人民邮出版社, 2003,10-13
[2] Nayar, S.K.Narasimhan, S.G.Vision in bad weather. Inproeeedings of the 7th International Conference on Computer Vision. 1999, 128-134
[3] S.G.Narasimhan, C.Wang, S.K.Nayar. All the images of an outdoor scene. In Proc.ECCV, 2002, 576-580
[4]杨晨.雾天图像增强算法研究: [南京理工大学硕士学位论文].南京:南京理工大学, 2007, 60-65
[5]王多超.图像去雾算法及其应用研究: [安徽大学硕士学位论文].安徽:安徽大学, 2010,12-19
[6] Bennett E P, McMillan L. Video Enhancement using Per-Pixel Virtual Exposures. ACM Trans.Graph, 2005, 24(3): 845-852
[7]何斌,马天予等. Visual C++数字图像处理.北京:人民邮电出版社, 2002,1-5
[8] Rafael C G, Richard E W.数字图像处理.北京:电子工业出版社, 2007,26-35
[9] Ted J C, Farhan A B. Analysis and Extensions of the Frankle_Mccann Retinex Algorithm.Journal of Electronic Imaging, 2004,13(l): 85-92
[10] Brian F, Forian C, John M. Retinex in Matlab. Journal of Electronic Imaging, 2004, 13(l): 48-57
[11] Florian C, Brian F. Tuning Retinex Pararneters. Journal of Electronic Imaging, 2004,13(l): 58-64
[12] Rahman Zia-Ur, Daniel J J, Glenn A W. Multiseale Retinex for Color Image Enhancement. In: Proc of IEEE Int Conf on Image Processing. 1996,20-24
[13]朱虹.大雾天气下图像的清晰化方法: [西安理工大学硕士学位论文].西安:西安理工大学, 2006, 18-30
[14]贾明桥.大雾天气下图像的清晰化方法: [西安理工大学硕士学位论文].西安:西安理工大学, 2006, 35-40
[15] L.Pirodda. Enhancing Visibility Through Fog. Optics & Laser Technology, 1997, 29(No.6): 293-299
[16]祝培.恶劣天气环境下图像的清晰化: [西安理工大学硕士学位论文].西安:西安理工大学, 2004, 36-42
[17]朱凯军,周焰,兰祖送.基于区域分割的雾天图像增强算法.计算机测量与控制, 2006, 14(5): 661-663
[18]祝培,朱虹,钱学明等.一种有雾天气图像景物影像的清晰化方法.中国图像图像学报, 2004, 9(1): 124-128
[19]孙奕敏.灾害性浓雾.北京:气象出版社, 1994, 1-4
[20]何俊,葛红,王玉峰.图像分割算法研究综述.计算机工程与科学, 2009, 31(12): 58-61
[21]周吕雄,于盛林.基于最小方差Snake模型的医学图像分割.生物医学工程杂志, 2007, 24(1): 32-35
[22]张建伟,罗剑,夏德深.一种基于遗传算法的双T_snake模型图像分割方法.中国图像图形学报, 2005, l0(1): 38-42
[23] Udupa J K, Wei L, Samarasekera S, etal. Multiple Sclerosis Lesion Qualification Using Fuzzy-Connected Principle. IEEE Trans on Med Imaging.1997,16(5): 598
[24] You Jianjie, zhou Zeming, Pheng Ann Heng, et a1. Simulated Annealing Based Simplified Snakes for Weak Edge Medical Image Segmentation. Journal of Image and Graphics, 2004, 9(1): 11-17
[25]陆剑锋,林海,潘志庚.白适应区域生长算法在医学图像分割中的应用.计算机辅助设计与图形学学报, 2005, 17(10): 2168-2173
[26] N.R.Pal and S.K.Pal. A review on image segmentation techniques.Pattern Reconition, 1993,16-22
[27] Rudolph G. Convergence analysis of canonical genetic algorithms. In: IEEE Trans on Neural Networks. 1994,18-22
[28] Bhanu B, Lee S, Ming J. Adaptive Image Segmentation using a Genetic Algorithm. IEEE Transaction on Systems, Man and Cybermetics, 1995,16-19
[29]黄菲,基于遗传算法的图像分割: [武汉科技大学硕士学位论文].武汉:武汉科技大学, 2008, 35-45
[30]李辉,基于改进遗传算法的图像分割: [东北师范大学硕士学位论文].长春:东北师范大学, 2004,38-44
[31]罗西平,田捷.图像分割方法综述.模式识别与人工智能, 1999, 9(3): 300-312
[32] OTSU N. A threshold selection method from gray level histogram. IEEE Transon Systems. Man and Cybernetics, 1979, 9 (1): 62-66
[33] KITTLER J, LL NGWORTH J. Minimum error thresholding. Pattern Recognit- ion, 1986, 19 (1): 41-47
[34] DUNN SM, HARWOOD D, DAVISL S. Local estimation of the uniform error threshold. IEEE Transon Pattern Analysis and Machine Intelligence, 1984, 6(6): 742-745
[35]陈果.图像阈值分割的Fisher准则函数法.仪器仪表学报,2003,24(6):564-567
[36]王培珍,杜培明.一种用于多阈值图像自动分割的混合遗传算法.中国图象图形学报, 2000, 5(6):66-69
[37]王月兰,曾迎生.信息融合技术在彩色图像分割方法中的应用.计算机学报, 2000, 7(2):55-60
[38]何文浩.基于改进遗传算法的图像分割技术研究: [武汉理工大学硕士学位论文].武汉:武汉理工大学, 2008,36-44
[39]徐小慧,张安.基于粒子群优化算法的最佳嫡阂值图像分割.计算机工程与应用, 2006, 10(11): 8-11
[40]郑宏,潘励.基于遗传算法的图像阈值的自动选取.中国图象图形学报,1999,4:1-9
[41]黄菲,基于遗传算法的图像分割: [武汉科技大学硕士学位论文].武汉:武汉科技大学, 2008,20-25
[42]张怀柱,向长波,宋建中.改进的遗传算法在实时图像分割中的应用.光学精密工程, 2008, 16(2): 333-337
[43]郝保明,包晓敏,汪亚明等.基于改进遗传算法的图像分割.浙江理工大学学报, 2008, 25(6): 700-703
[44]龚纯,王正林.精通MATLAB最优化计算.北京:电子工业出版社, 2009,100-120
[45]张金龙,赵芙生.基于遗传算法的三维重构图像阈值分割.南京师范大学学报, 2005, 5 (1): 5-7
[46]李康顺,李茂民,张文生.一种基于改进遗传算法的图像分割方法.计算机应用研究, 2009, 26(11): 4364-4367
[47]章毓晋.图像工程(上册)图像处理.第2版.北京:清华大学出版社, 2006,10-20
[48]贾明桥.大雾天气下图像的清晰化方法: [西安理工大学硕士学位论文].西安:西安理工大学, 2006,40-48
[49] Kenneth R. Castleman.数字图像处理.北京:电子工业出版社. 2002, 75-76
[50] Kim J Y, Kim L S, Hwang S H. An Advanced Contrast Enhancement Using Partially Overlapped Sub-block Histogram Equalization. IEEE ISCAS 2000 Geneva, Switzerland.2000-05
[51]杨词银,黄廉卿.基于幂函数的加权自适应直方图均衡.光电子·激光, 2002, 13(5): 515-517
[52]刘瑞剑.低能见度条件下图像清晰化处理研究: [中北大学硕士学位论文].广州:中北大学, 2008
[53] E.J.McCartney. Optics of the Atmosphere: Scattering by molecules and particles. John Wiley and Sons,1975,10-15
[2] Nayar, S.K.Narasimhan, S.G.Vision in bad weather. Inproeeedings of the 7th International Conference on Computer Vision. 1999, 128-134
[3] S.G.Narasimhan, C.Wang, S.K.Nayar. All the images of an outdoor scene. In Proc.ECCV, 2002, 576-580
[4]杨晨.雾天图像增强算法研究: [南京理工大学硕士学位论文].南京:南京理工大学, 2007, 60-65
[5]王多超.图像去雾算法及其应用研究: [安徽大学硕士学位论文].安徽:安徽大学, 2010,12-19
[6] Bennett E P, McMillan L. Video Enhancement using Per-Pixel Virtual Exposures. ACM Trans.Graph, 2005, 24(3): 845-852
[7]何斌,马天予等. Visual C++数字图像处理.北京:人民邮电出版社, 2002,1-5
[8] Rafael C G, Richard E W.数字图像处理.北京:电子工业出版社, 2007,26-35
[9] Ted J C, Farhan A B. Analysis and Extensions of the Frankle_Mccann Retinex Algorithm.Journal of Electronic Imaging, 2004,13(l): 85-92
[10] Brian F, Forian C, John M. Retinex in Matlab. Journal of Electronic Imaging, 2004, 13(l): 48-57
[11] Florian C, Brian F. Tuning Retinex Pararneters. Journal of Electronic Imaging, 2004,13(l): 58-64
[12] Rahman Zia-Ur, Daniel J J, Glenn A W. Multiseale Retinex for Color Image Enhancement. In: Proc of IEEE Int Conf on Image Processing. 1996,20-24
[13]朱虹.大雾天气下图像的清晰化方法: [西安理工大学硕士学位论文].西安:西安理工大学, 2006, 18-30
[14]贾明桥.大雾天气下图像的清晰化方法: [西安理工大学硕士学位论文].西安:西安理工大学, 2006, 35-40
[15] L.Pirodda. Enhancing Visibility Through Fog. Optics & Laser Technology, 1997, 29(No.6): 293-299
[16]祝培.恶劣天气环境下图像的清晰化: [西安理工大学硕士学位论文].西安:西安理工大学, 2004, 36-42
[17]朱凯军,周焰,兰祖送.基于区域分割的雾天图像增强算法.计算机测量与控制, 2006, 14(5): 661-663
[18]祝培,朱虹,钱学明等.一种有雾天气图像景物影像的清晰化方法.中国图像图像学报, 2004, 9(1): 124-128
[19]孙奕敏.灾害性浓雾.北京:气象出版社, 1994, 1-4
[20]何俊,葛红,王玉峰.图像分割算法研究综述.计算机工程与科学, 2009, 31(12): 58-61
[21]周吕雄,于盛林.基于最小方差Snake模型的医学图像分割.生物医学工程杂志, 2007, 24(1): 32-35
[22]张建伟,罗剑,夏德深.一种基于遗传算法的双T_snake模型图像分割方法.中国图像图形学报, 2005, l0(1): 38-42
[23] Udupa J K, Wei L, Samarasekera S, etal. Multiple Sclerosis Lesion Qualification Using Fuzzy-Connected Principle. IEEE Trans on Med Imaging.1997,16(5): 598
[24] You Jianjie, zhou Zeming, Pheng Ann Heng, et a1. Simulated Annealing Based Simplified Snakes for Weak Edge Medical Image Segmentation. Journal of Image and Graphics, 2004, 9(1): 11-17
[25]陆剑锋,林海,潘志庚.白适应区域生长算法在医学图像分割中的应用.计算机辅助设计与图形学学报, 2005, 17(10): 2168-2173
[26] N.R.Pal and S.K.Pal. A review on image segmentation techniques.Pattern Reconition, 1993,16-22
[27] Rudolph G. Convergence analysis of canonical genetic algorithms. In: IEEE Trans on Neural Networks. 1994,18-22
[28] Bhanu B, Lee S, Ming J. Adaptive Image Segmentation using a Genetic Algorithm. IEEE Transaction on Systems, Man and Cybermetics, 1995,16-19
[29]黄菲,基于遗传算法的图像分割: [武汉科技大学硕士学位论文].武汉:武汉科技大学, 2008, 35-45
[30]李辉,基于改进遗传算法的图像分割: [东北师范大学硕士学位论文].长春:东北师范大学, 2004,38-44
[31]罗西平,田捷.图像分割方法综述.模式识别与人工智能, 1999, 9(3): 300-312
[32] OTSU N. A threshold selection method from gray level histogram. IEEE Transon Systems. Man and Cybernetics, 1979, 9 (1): 62-66
[33] KITTLER J, LL NGWORTH J. Minimum error thresholding. Pattern Recognit- ion, 1986, 19 (1): 41-47
[34] DUNN SM, HARWOOD D, DAVISL S. Local estimation of the uniform error threshold. IEEE Transon Pattern Analysis and Machine Intelligence, 1984, 6(6): 742-745
[35]陈果.图像阈值分割的Fisher准则函数法.仪器仪表学报,2003,24(6):564-567
[36]王培珍,杜培明.一种用于多阈值图像自动分割的混合遗传算法.中国图象图形学报, 2000, 5(6):66-69
[37]王月兰,曾迎生.信息融合技术在彩色图像分割方法中的应用.计算机学报, 2000, 7(2):55-60
[38]何文浩.基于改进遗传算法的图像分割技术研究: [武汉理工大学硕士学位论文].武汉:武汉理工大学, 2008,36-44
[39]徐小慧,张安.基于粒子群优化算法的最佳嫡阂值图像分割.计算机工程与应用, 2006, 10(11): 8-11
[40]郑宏,潘励.基于遗传算法的图像阈值的自动选取.中国图象图形学报,1999,4:1-9
[41]黄菲,基于遗传算法的图像分割: [武汉科技大学硕士学位论文].武汉:武汉科技大学, 2008,20-25
[42]张怀柱,向长波,宋建中.改进的遗传算法在实时图像分割中的应用.光学精密工程, 2008, 16(2): 333-337
[43]郝保明,包晓敏,汪亚明等.基于改进遗传算法的图像分割.浙江理工大学学报, 2008, 25(6): 700-703
[44]龚纯,王正林.精通MATLAB最优化计算.北京:电子工业出版社, 2009,100-120
[45]张金龙,赵芙生.基于遗传算法的三维重构图像阈值分割.南京师范大学学报, 2005, 5 (1): 5-7
[46]李康顺,李茂民,张文生.一种基于改进遗传算法的图像分割方法.计算机应用研究, 2009, 26(11): 4364-4367
[47]章毓晋.图像工程(上册)图像处理.第2版.北京:清华大学出版社, 2006,10-20
[48]贾明桥.大雾天气下图像的清晰化方法: [西安理工大学硕士学位论文].西安:西安理工大学, 2006,40-48
[49] Kenneth R. Castleman.数字图像处理.北京:电子工业出版社. 2002, 75-76
[50] Kim J Y, Kim L S, Hwang S H. An Advanced Contrast Enhancement Using Partially Overlapped Sub-block Histogram Equalization. IEEE ISCAS 2000 Geneva, Switzerland.2000-05
[51]杨词银,黄廉卿.基于幂函数的加权自适应直方图均衡.光电子·激光, 2002, 13(5): 515-517
[52]刘瑞剑.低能见度条件下图像清晰化处理研究: [中北大学硕士学位论文].广州:中北大学, 2008
[53] E.J.McCartney. Optics of the Atmosphere: Scattering by molecules and particles. John Wiley and Sons,1975,10-15