印版显微灰度图像二值化算法研究
|
摘要
网点增大是印刷品质量的主要控制因素之一,而网点面积覆盖率是计算网点增大的基础。长期以来印版网点面积率的检测一直是印刷复制过程中检测技术的难点,以至于业界只得将印品和胶片上的网点面积率之差定义为印刷过程中的网点扩大值。随着直接制版技术(CTP)的快速发展和在实际生产中广泛的应用,直接测量印版的网点面积率成了CTP过程中质量控制的关键,同时也是对印刷品质量进行有效控制的基础。
针对印版图像的反差小、网点边缘模糊的特点,以及暗调和亮调层次的网点显微图像难于分割的情况,本课题在通用的图像分割算法中,有目的地选用善长于分割一般模糊图像的最大类间方差法、模糊c均值聚类算法(FCM)和脉冲耦合神经网络(PCNN)算法等三种方法,进行印版图像分割试验,以国际印刷行业认可的iCPlateⅡ印版检测仪测量数据作为测量标准,研究适于印版显微图像的分割算法。
文中在最大类间方差法原始算法基础上,引进基于二维直方图的最大类间方差算法,并用一次和二次高斯平滑滤波值构成的二维直方图代替由像素灰度值和像素的邻域灰度均值构成的二维直方图,建立了基于改进二维直方图的最大类间方差法,较好地实现对印版图像进行自适应分割。
研究中使用高斯平滑滤波器取代均值滤波器,改进了基于二维直方图加权的FCM算法。首次引入了平滑因子m的自适应算法。同时采用特定的预处理方法,建立了适于印版图像的基于改进二维直方图加权及平滑因子m自适应的FCM算法。
本课题首次将PCNN算法引入到印版显微网点图像的分割中。为了更好地适应印版图像分割的应用要求,研究中对基本模型进行了简化和改进,并且摒弃了常用的收敛准则,结合点火频率思想建立了一种基于点火频率的改进脉冲耦合神经网络图像分割新方法。
在课题研究中对三种典型算法分别作了改进,取得了令人满意的结果,获得了能够测量小反差印版图像并具有较高精度的印版网点图像自适应二值化分割算法,为开发出具有自主知识产权的印版网点图像检测仪进行了有实际意义的探索。
Dot area coverage is the foundation of computing dot gain which is one of the major factors to control quality of printing. For long time, due to the difficulty of accuracy detection of dot area coverage on the plate image, only the dot aera coverage difference between printing material and film has been adopted by the industry instead of the real dot gain. With rapid development of the Computer-To-Plate (CTP) technology, the direct detection of dot area coverage on the plate become one of the key technologies to control quality during printing and copy processes.
With regards to the characteristic of low contrast in plate image and fuzzy dot edge, three segmentation algorithms were applied in this paper, which are good at processing fuzzy images, to find the segmentation algorithms which are suitable for plate microscopic image. They are the Maximum Between-Cluster Variance algorithm, the Fuzzy C-Means clustering algorithm and the Pulse Coupled Neural Network algorithm. During the research, detecion data collected by iCPlate II which is recognized by international printing industry as detection standards.
During the research, The Maximum Between-Cluster Variance algorithm based on 2D histogram is introduced. Moreover, an improved Maximum Between-Cluster Variance algorithm on the basis of the improved 2D histogram composed of the result of once and twice Guassian filtering is built up. It is proved by experiments that new algorithm can be used as an automatic segmentation algorithm for plate microscopic images.
The Fuzzy C-Means clustering algorithm is improved by using a modified weighted 2D histograms, introduing an adaptive computaion of smoothing factor m which is the first time applied in processing plate image and obtaining a special pre-processing method designed for plate microscopic image.
For Pulse Coupled Neural Network (PCNN) algorithm, it is the first time to be utilized for plate images. In order to better meet application requirement of image segmentation, the basic neuromime of PCNN is simplified and improved. At the same time, the segmentation method based on fired frequency is implemented instead of using common convergence criteria.
The research improves three algorithms and attains satisfactory results. These adaptive binarization algorithms for plate image can produce dot image with higher accuracy with low contrast. The research explores realistically for the purpose of developing plate dot image detecting instrument with independent intellectual property rights.
针对印版图像的反差小、网点边缘模糊的特点,以及暗调和亮调层次的网点显微图像难于分割的情况,本课题在通用的图像分割算法中,有目的地选用善长于分割一般模糊图像的最大类间方差法、模糊c均值聚类算法(FCM)和脉冲耦合神经网络(PCNN)算法等三种方法,进行印版图像分割试验,以国际印刷行业认可的iCPlateⅡ印版检测仪测量数据作为测量标准,研究适于印版显微图像的分割算法。
文中在最大类间方差法原始算法基础上,引进基于二维直方图的最大类间方差算法,并用一次和二次高斯平滑滤波值构成的二维直方图代替由像素灰度值和像素的邻域灰度均值构成的二维直方图,建立了基于改进二维直方图的最大类间方差法,较好地实现对印版图像进行自适应分割。
研究中使用高斯平滑滤波器取代均值滤波器,改进了基于二维直方图加权的FCM算法。首次引入了平滑因子m的自适应算法。同时采用特定的预处理方法,建立了适于印版图像的基于改进二维直方图加权及平滑因子m自适应的FCM算法。
本课题首次将PCNN算法引入到印版显微网点图像的分割中。为了更好地适应印版图像分割的应用要求,研究中对基本模型进行了简化和改进,并且摒弃了常用的收敛准则,结合点火频率思想建立了一种基于点火频率的改进脉冲耦合神经网络图像分割新方法。
在课题研究中对三种典型算法分别作了改进,取得了令人满意的结果,获得了能够测量小反差印版图像并具有较高精度的印版网点图像自适应二值化分割算法,为开发出具有自主知识产权的印版网点图像检测仪进行了有实际意义的探索。
Dot area coverage is the foundation of computing dot gain which is one of the major factors to control quality of printing. For long time, due to the difficulty of accuracy detection of dot area coverage on the plate image, only the dot aera coverage difference between printing material and film has been adopted by the industry instead of the real dot gain. With rapid development of the Computer-To-Plate (CTP) technology, the direct detection of dot area coverage on the plate become one of the key technologies to control quality during printing and copy processes.
With regards to the characteristic of low contrast in plate image and fuzzy dot edge, three segmentation algorithms were applied in this paper, which are good at processing fuzzy images, to find the segmentation algorithms which are suitable for plate microscopic image. They are the Maximum Between-Cluster Variance algorithm, the Fuzzy C-Means clustering algorithm and the Pulse Coupled Neural Network algorithm. During the research, detecion data collected by iCPlate II which is recognized by international printing industry as detection standards.
During the research, The Maximum Between-Cluster Variance algorithm based on 2D histogram is introduced. Moreover, an improved Maximum Between-Cluster Variance algorithm on the basis of the improved 2D histogram composed of the result of once and twice Guassian filtering is built up. It is proved by experiments that new algorithm can be used as an automatic segmentation algorithm for plate microscopic images.
The Fuzzy C-Means clustering algorithm is improved by using a modified weighted 2D histograms, introduing an adaptive computaion of smoothing factor m which is the first time applied in processing plate image and obtaining a special pre-processing method designed for plate microscopic image.
For Pulse Coupled Neural Network (PCNN) algorithm, it is the first time to be utilized for plate images. In order to better meet application requirement of image segmentation, the basic neuromime of PCNN is simplified and improved. At the same time, the segmentation method based on fired frequency is implemented instead of using common convergence criteria.
The research improves three algorithms and attains satisfactory results. These adaptive binarization algorithms for plate image can produce dot image with higher accuracy with low contrast. The research explores realistically for the purpose of developing plate dot image detecting instrument with independent intellectual property rights.
引文
[1]徐锦林.印刷工程导论[M].化学工业出版社,2006：148-149
[2]贺文琼,陈亚维.数码印刷机的输出分辨力特性分析[J].数码印刷,2003,(4)：14
[3]David R.Wyble, Roy S.Berns. A Critical Review of Spectral Models Applied to Binary Color Printing[J]. Color Research and Application[J],1999,25(1):4-19
[4]姚海根.数字加网技术[M].北京：印刷工业出版社,2000：22
[5]张逸新.分色制版技术[M].北京：中国轻工业出版社,2001：138
[6]胡涛,付兆敏,吴学毅.基于数字图像的彩色印品网点面积检测装置的研究[J].印刷杂志,2003,10：57
[7]胡涛,郑敏东,张晓惠.彩色印刷品质量检测技术的发展[J].印刷世界,2004,(7)：17-18
[8]马步庄.印刷网点图像显微检测仪系统开发与检测算法研究[D].大连：大连轻工业学院,2006：
[9]彭策.印刷品质量控制[M].化学工业出版社,2004：35-42
[10]张仁英.用EQPM法测量网点面积.今日印刷,1999,(1)：118-119
[11]蒋磊.基于图像分割处理的印版检测技术的研究[D].西安：西安理工大学,2004
[12]王小红,衣永政.基于网点图像模糊性的图像分割方法的研究[J].印刷工程,2005,(3)：21-25
[13]周春霞,魏敏,唐正宁.基于数学形态学的印刷网点图像分割方法[J].包装工程,2007,(6)：14-15
[14]潘喆,吴一全.二维指数熵图像阈值选取方法及其快速算法[J].计算机应用,2007,(4)：982-985
[15]Fu K S, Mui J K. A Survey of Threshold Selection Techniques [J]. Pattern Recognition,1981, (13):3-16
[16]吴一全,朱兆达.图像处理中阈值选取方法30年(1962-1992)的进展(一)[J].数据采集与处理,1993,8(3)：193-201
[17]马义德,李廉,王亚馥,戴若兰.脉冲耦合神经网络原理及其应用[M].科学出版社,2006：8
[18]Otsu N. Disciminant and Lesat Square Threshold Selection[R]. In Proc 4IJCPR,1978:592-596
[19]Otsu N. A Threshold Selection Method From Gray-level Histogram[J]. IEEE Trans on SMC-9,1979:62-66
[20]付忠良.图像阈值选取方法-Otsu方法的推广[J].计算机应用,2000,(5)：37-39
[21]刘建庄,栗文清.灰度图像的二维Otsu自动闽值分割法[J].自动化学报,1993,19(1)：101-105
[22]汪海洋,潘德炉,夏德深.二维Otsu自适应阈值选取算法的快速实现[J].自动化学报2007,(9)：968-971
[23]李改梅,杨润玲,周军妮.基于二维直方图与FCM相结合的图像快速分割方法[M].电子技术应用,2007,(15)：176-178
[24]杨润玲.基于FCM类型算法的自动图像分割方法研究[D].西安：西安电子科技大学,2006：7
[25]Ruspini E H. A new approach to clustering. Inf.Cont.,1969,15(1):22-32
[26]Dunn J C. A graph theoretic analysis of pattern classification via Tamura's fuzzy relation [M]. IEEE Trans. SMC,1974,4(3):310-313
[27]Bezdek J C. Pattern Recognition With Fuzzy Objective Function Algorithms[M]. Plenum Press,1981
[28]高新波.模糊聚类分析及其应用[M].西安电子科技大学出版社,2004：109
[29]Bezdek J C. A Physical Interpretation of Fuzzy ISODATA[J]. IEEE Trans.SMC,1976,6(3):387-390
[30]Chan K P, Cheung Y S. Clustering of Clusters. Pattern Recogintion,1992,25(2):211-217
[31]Bezek J C, Hathaway R J, et al. Convergence and Theory for Fuzzy C-means Clustering Algorithm:counterexamples and repaires[J]. IEEE Trans.PAMI,1987,17(5):873-877
[32]Pal N R, Bezek J C. On Clustering Validation for the Fuzzy C-means Model[J]. IEEE Trans. FS, 1995,3(3):370-379
[33]高新波,裴继红,谢维信.模糊C均值聚类算法中加权指数m的研究[J].电子学报,2000,28(4)：80-84
[34]宫改云,高新波,伍忠东.FCM聚类算法中模糊加权指数m的优选方法.模糊数学与教学.2005,19(1)：143-148
[35]王思贤,曾发龙.平滑图像噪声的差值滤波法[J].电子科学学刊,2000,22(3)：411-414.
[36]Rafael C. Gonzalez, Richard E.Woods, Steven L Eddins.数字图像处理(MATLAB版).电子工业出版社[M].2005：48
[37]葛哲学,孙志强.神经网络理论与MATLAB R2007实现[M].电子工业出版社,2007：1
[38]R Eckhorn, H J Reitboeck, M Arndt, et al. Feature Linking Via Synchronization Among Distributed Assemblies:Simulation of Results From Cat Cortex[J]. Neural Comput.,1990,2(3):293-307
[39]R Eckhorn, A Frien, R Bauer, et al. High frequency oscillations in primary visual cortex of awake monkey[J]. Neuroreport,1999,4(3):243-246
[40]马义德,李廉,绽琨,王兆滨.脉冲耦合神经网络与数字图像处理[M].科学出版社,2008：8
[41]刘勃,马义德,钱志伯.一种基于交叉熵的改进型PCNN图像自动分割新方法[J].中国图像图形学报,2005,10(5)：579-584
[42]马义德, 戴若兰,李廉.一种基于脉冲耦合神经网络和图像熵的自动图像分割方法[J].通信学报,2002,23(1)：46-51
[43]黄荣杰,崔世林.基于点火频率的PCNN灰度图像分割方法研究[J].图像与多媒体技术,2007,64-66
[2]贺文琼,陈亚维.数码印刷机的输出分辨力特性分析[J].数码印刷,2003,(4)：14
[3]David R.Wyble, Roy S.Berns. A Critical Review of Spectral Models Applied to Binary Color Printing[J]. Color Research and Application[J],1999,25(1):4-19
[4]姚海根.数字加网技术[M].北京：印刷工业出版社,2000：22
[5]张逸新.分色制版技术[M].北京：中国轻工业出版社,2001：138
[6]胡涛,付兆敏,吴学毅.基于数字图像的彩色印品网点面积检测装置的研究[J].印刷杂志,2003,10：57
[7]胡涛,郑敏东,张晓惠.彩色印刷品质量检测技术的发展[J].印刷世界,2004,(7)：17-18
[8]马步庄.印刷网点图像显微检测仪系统开发与检测算法研究[D].大连：大连轻工业学院,2006：
[9]彭策.印刷品质量控制[M].化学工业出版社,2004：35-42
[10]张仁英.用EQPM法测量网点面积.今日印刷,1999,(1)：118-119
[11]蒋磊.基于图像分割处理的印版检测技术的研究[D].西安：西安理工大学,2004
[12]王小红,衣永政.基于网点图像模糊性的图像分割方法的研究[J].印刷工程,2005,(3)：21-25
[13]周春霞,魏敏,唐正宁.基于数学形态学的印刷网点图像分割方法[J].包装工程,2007,(6)：14-15
[14]潘喆,吴一全.二维指数熵图像阈值选取方法及其快速算法[J].计算机应用,2007,(4)：982-985
[15]Fu K S, Mui J K. A Survey of Threshold Selection Techniques [J]. Pattern Recognition,1981, (13):3-16
[16]吴一全,朱兆达.图像处理中阈值选取方法30年(1962-1992)的进展(一)[J].数据采集与处理,1993,8(3)：193-201
[17]马义德,李廉,王亚馥,戴若兰.脉冲耦合神经网络原理及其应用[M].科学出版社,2006：8
[18]Otsu N. Disciminant and Lesat Square Threshold Selection[R]. In Proc 4IJCPR,1978:592-596
[19]Otsu N. A Threshold Selection Method From Gray-level Histogram[J]. IEEE Trans on SMC-9,1979:62-66
[20]付忠良.图像阈值选取方法-Otsu方法的推广[J].计算机应用,2000,(5)：37-39
[21]刘建庄,栗文清.灰度图像的二维Otsu自动闽值分割法[J].自动化学报,1993,19(1)：101-105
[22]汪海洋,潘德炉,夏德深.二维Otsu自适应阈值选取算法的快速实现[J].自动化学报2007,(9)：968-971
[23]李改梅,杨润玲,周军妮.基于二维直方图与FCM相结合的图像快速分割方法[M].电子技术应用,2007,(15)：176-178
[24]杨润玲.基于FCM类型算法的自动图像分割方法研究[D].西安：西安电子科技大学,2006：7
[25]Ruspini E H. A new approach to clustering. Inf.Cont.,1969,15(1):22-32
[26]Dunn J C. A graph theoretic analysis of pattern classification via Tamura's fuzzy relation [M]. IEEE Trans. SMC,1974,4(3):310-313
[27]Bezdek J C. Pattern Recognition With Fuzzy Objective Function Algorithms[M]. Plenum Press,1981
[28]高新波.模糊聚类分析及其应用[M].西安电子科技大学出版社,2004：109
[29]Bezdek J C. A Physical Interpretation of Fuzzy ISODATA[J]. IEEE Trans.SMC,1976,6(3):387-390
[30]Chan K P, Cheung Y S. Clustering of Clusters. Pattern Recogintion,1992,25(2):211-217
[31]Bezek J C, Hathaway R J, et al. Convergence and Theory for Fuzzy C-means Clustering Algorithm:counterexamples and repaires[J]. IEEE Trans.PAMI,1987,17(5):873-877
[32]Pal N R, Bezek J C. On Clustering Validation for the Fuzzy C-means Model[J]. IEEE Trans. FS, 1995,3(3):370-379
[33]高新波,裴继红,谢维信.模糊C均值聚类算法中加权指数m的研究[J].电子学报,2000,28(4)：80-84
[34]宫改云,高新波,伍忠东.FCM聚类算法中模糊加权指数m的优选方法.模糊数学与教学.2005,19(1)：143-148
[35]王思贤,曾发龙.平滑图像噪声的差值滤波法[J].电子科学学刊,2000,22(3)：411-414.
[36]Rafael C. Gonzalez, Richard E.Woods, Steven L Eddins.数字图像处理(MATLAB版).电子工业出版社[M].2005：48
[37]葛哲学,孙志强.神经网络理论与MATLAB R2007实现[M].电子工业出版社,2007：1
[38]R Eckhorn, H J Reitboeck, M Arndt, et al. Feature Linking Via Synchronization Among Distributed Assemblies:Simulation of Results From Cat Cortex[J]. Neural Comput.,1990,2(3):293-307
[39]R Eckhorn, A Frien, R Bauer, et al. High frequency oscillations in primary visual cortex of awake monkey[J]. Neuroreport,1999,4(3):243-246
[40]马义德,李廉,绽琨,王兆滨.脉冲耦合神经网络与数字图像处理[M].科学出版社,2008：8
[41]刘勃,马义德,钱志伯.一种基于交叉熵的改进型PCNN图像自动分割新方法[J].中国图像图形学报,2005,10(5)：579-584
[42]马义德, 戴若兰,李廉.一种基于脉冲耦合神经网络和图像熵的自动图像分割方法[J].通信学报,2002,23(1)：46-51
[43]黄荣杰,崔世林.基于点火频率的PCNN灰度图像分割方法研究[J].图像与多媒体技术,2007,64-66
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |