灰度图像的边缘检测研究
|
摘要
边缘检测是图像处理与分析中最基础的内容之一,也是至今仍没有得到圆满解决的一类问题。图像的边缘包含了图像的位置、轮廓等特征,是图像的基本特征之一,广泛地应用于特征描述、图像分割、图像增强、模式识别等图像分析和处理中。因此,图像边缘的检测方法,一直是图像处理与分析技术中的研究热点。然而,至今发表的有关边缘检测的理论和方法尚存在许多不足之处,比如在检测精度和去噪方面很难达到令人满意的效果。
本文针对边缘检测中存在的问题,在对一些传统和新兴的边缘检测方法进行归纳的基础上,围绕灰度图像的边缘检测做了如下一些创新性和探索性工作:
研究和分析了常用的图像滤波方法,在此基础上设计了一种将改进中值滤波和自适应滤波相结合的二阶段滤波方法。其中,改进的中值滤波使用多个模板自适应地滤波,以此减少滤除椒盐噪声过程中对图像的损坏,而自适应滤波则能在滤除高斯噪声的过程中很好的保护图像边缘。
详细研究了Sobel算子和Laplacian算子在边缘检测中的特点,并根据这些特点设计了两种结合Sobel算子和Laplacian算子的边缘检测方法,即基于梯度相乘的边缘检测算法和基于边缘分类的边缘检测算法。
研究了边缘处灰度变化的特点,根据边缘处在图像灰度对比较大的地方这一特点定义了局部灰度差异,而根据沿着边缘方向灰度变化较小,同时沿着垂直于边缘的方向灰度变化较大的特点定义了边缘似然度。在此基础上设计了基于局部灰度差异的边缘检测和基于边缘似然度的边缘检测这两种基于模板思想的边缘检测方法。
详细研究了已有一些模糊增强算法的缺陷,并分别从运算量、增强变换的选择、渡越点的选择以及增强变换的调整方面给出了改进意见,最后设计了一种基于改进模糊增强的边缘检测算法,并使用双阈值检测的方法来保障所检测边缘的可靠性和完整性。
Edge detection is one of the most fundamental aspects in image processing and analyzing,which is still unsolved.Image’s edges include image’s features such as position and outline,which is the fundamental features of the image. Edge detection is widely used in image analysis and processing such as feature description, image segmentation, image enhancement and pattern recognition etc. so edge detection is the research hot spot in the technology of image processing and analysis. In recent years, many theories and technologies on edge detection are presented. The theories and methods on edge detection still have much defect to be improved. For instance in the detection accuracy and noise removement aspects,these theories and methods are not satisfactory.
In this thesis,some traditional and emerging edge detection methods are being summarized.Then some innovative and exploratory work is being done on gray image edge detection as follows:
Commonly used image filtering methods are investigated and analysed in detail and a two-stage filtering method which combines improved median filtering and adaptive filtering.Serveral templates are adaptively used in improved median filtering and damage to the image is minimized during the process of eliminating salt and papper noise.Meanwhile, image edges are well protected during the process of eliminating gaussian noise by adaptive filtering.
Sobel operator’s characteristics and Laplacian operator’s characteristics in edge detection are investigated in detail.Then edge detection algorithm that based on gradient multiplication and edge detection algorithm that based on edge classification are designed.They are two edge detection algorithms that combine Sobel operator and Laplacian operator.
Characteristic of gray-scale change of the edge is researched.According to the feature that the edges are mainly locate at the position where gray contrast is large and local differences in gray is defined.According to the feature that the gray difference change along the orientation of the edge is small while the gray difference change is significant along the orientation that perpendicular to the edge and edge likelihood is defined. Finally edge detection algorithm that based on local gray difference and edge detection algorithm that based on edge likelihood is defined.These are two template-based edge detection algorithms.
Defects of some existing fuzzy enhancement algorithms are investigated in detail. Then Improvements on computation, selection of enhancement transformation, selection of crossover point and adjustment of enhancement transformation are proposed.Finally an edge detection algorithm that based on improved fuzzy enhancement algorithm is designed.In order to guarantee the reliability and integrity of the edge that we detect,dual-threshold is used.
本文针对边缘检测中存在的问题,在对一些传统和新兴的边缘检测方法进行归纳的基础上,围绕灰度图像的边缘检测做了如下一些创新性和探索性工作:
研究和分析了常用的图像滤波方法,在此基础上设计了一种将改进中值滤波和自适应滤波相结合的二阶段滤波方法。其中,改进的中值滤波使用多个模板自适应地滤波,以此减少滤除椒盐噪声过程中对图像的损坏,而自适应滤波则能在滤除高斯噪声的过程中很好的保护图像边缘。
详细研究了Sobel算子和Laplacian算子在边缘检测中的特点,并根据这些特点设计了两种结合Sobel算子和Laplacian算子的边缘检测方法,即基于梯度相乘的边缘检测算法和基于边缘分类的边缘检测算法。
研究了边缘处灰度变化的特点,根据边缘处在图像灰度对比较大的地方这一特点定义了局部灰度差异,而根据沿着边缘方向灰度变化较小,同时沿着垂直于边缘的方向灰度变化较大的特点定义了边缘似然度。在此基础上设计了基于局部灰度差异的边缘检测和基于边缘似然度的边缘检测这两种基于模板思想的边缘检测方法。
详细研究了已有一些模糊增强算法的缺陷,并分别从运算量、增强变换的选择、渡越点的选择以及增强变换的调整方面给出了改进意见,最后设计了一种基于改进模糊增强的边缘检测算法,并使用双阈值检测的方法来保障所检测边缘的可靠性和完整性。
Edge detection is one of the most fundamental aspects in image processing and analyzing,which is still unsolved.Image’s edges include image’s features such as position and outline,which is the fundamental features of the image. Edge detection is widely used in image analysis and processing such as feature description, image segmentation, image enhancement and pattern recognition etc. so edge detection is the research hot spot in the technology of image processing and analysis. In recent years, many theories and technologies on edge detection are presented. The theories and methods on edge detection still have much defect to be improved. For instance in the detection accuracy and noise removement aspects,these theories and methods are not satisfactory.
In this thesis,some traditional and emerging edge detection methods are being summarized.Then some innovative and exploratory work is being done on gray image edge detection as follows:
Commonly used image filtering methods are investigated and analysed in detail and a two-stage filtering method which combines improved median filtering and adaptive filtering.Serveral templates are adaptively used in improved median filtering and damage to the image is minimized during the process of eliminating salt and papper noise.Meanwhile, image edges are well protected during the process of eliminating gaussian noise by adaptive filtering.
Sobel operator’s characteristics and Laplacian operator’s characteristics in edge detection are investigated in detail.Then edge detection algorithm that based on gradient multiplication and edge detection algorithm that based on edge classification are designed.They are two edge detection algorithms that combine Sobel operator and Laplacian operator.
Characteristic of gray-scale change of the edge is researched.According to the feature that the edges are mainly locate at the position where gray contrast is large and local differences in gray is defined.According to the feature that the gray difference change along the orientation of the edge is small while the gray difference change is significant along the orientation that perpendicular to the edge and edge likelihood is defined. Finally edge detection algorithm that based on local gray difference and edge detection algorithm that based on edge likelihood is defined.These are two template-based edge detection algorithms.
Defects of some existing fuzzy enhancement algorithms are investigated in detail. Then Improvements on computation, selection of enhancement transformation, selection of crossover point and adjustment of enhancement transformation are proposed.Finally an edge detection algorithm that based on improved fuzzy enhancement algorithm is designed.In order to guarantee the reliability and integrity of the edge that we detect,dual-threshold is used.
引文
[1]章毓晋.图像分析(第二版).图像工程.北京:清华大学出版社,2005,75-76
[2] Lee J,Haralick R, Shapiro L.Morphologic edge detection.IEEE Transactions on Robotics and Automation,1987,3(2):142-156
[3]甘金来,刘钊.图像边缘检测算法的比较研究:[硕士学位论文].成都:电子科技大学,2005
[4]冈萨雷斯.数字图像处理(第二版),阮秋琦,阮宇智等.北京:电子工业出版社.2003,464
[5]巴本冬,吴晓娟等.一种快速医学图像亚像素边缘检测方法.山东大学学报(工学版),2005,35(5):63-67
[6]孙红光,李勇.基于双正交小波弱目标边缘提取方法.长春理工大学报,2004,27(2):53-55
[7]周术诚,耿国华.基于自适应阈值的三维图像多尺度边缘检测.西北大学学报(自然科学版),2006,36(3):393-397
[8]金晟业,陈圣波等.遥感图像边缘提取微分方法及其应用.遥感技术与应用,2008,23(6):729-734
[9]谭海峰,赵文杰等.基于过渡区域模糊增强的合成孔径雷达图像边缘提取算法.计算机与现代化,2008,10:24-29
[10]李新,谢永智等.基于运动图像序列的啤酒瓶凸性字符快速提取方法.桂林工学院学报,2005,25(1):97-100
[11]王新,黄兆云.基于多结构元素的数学形态学图像边缘检测.计算机工程与应用,2008,44(7):89-112
[12] Yuqian Zhao,Weihua Gui.Edge Detection Based on Multi-Structure Elements Morphology.IEEE,2006:9795-9798
[13]孙达,刘家锋.基于概率密度梯度的边缘检测.计算机学报,2009,32(2),299-307
[14]冯会真,夏哲雷等.基于神经网络的图像边缘检测方法.浙江:中国计量学院学报,2006,17(4):289-291
[15]张东波,王耀南.FCM聚类算法和粗糙集在医疗图像分割中的应用.仪器仪表学报,2006,27(12):1683-1687
[16]刘曙光,刘明远等.基于Canny准则的基数B样条小波边缘检测.信号处理,2001,17(5):418-423
[17]虞伟,张小刚.遗传算法优化的形态学边缘检测算子.微计算子信息,2008,24(9):168-170
[18]梁光明,孙即祥等. Otsu算法在Canny算子中的应用.湖南:国防科技大学学报,2003,25(5):36-39
[19] Junaid I. Siddique. Kenneth E.BarnerWavelet-based multiresolution edge detection utilizing gray lever edge maps. IEEE,1998:550-554
[20]刘力双,张姚等.图像的快速亚像素边缘检测方法.光电子,2005,16(8),993-996
[21]苏红彬,白凤娥.边缘检测方法及其在医学图像处理中的应用研究.光电子技术与信息,2005,18(2):23-27
[22]薛丽霞,王佐成等.基于多维云空间的多光谱遥感影像边缘检测研究.测绘科学,2008,33(1):188-190
[23]王鹏,耿相铭.基于Prewitt算子的快速帧内编码算法.微计算机信息(测控自动化),2006,22(7-1):192-194
[24]李弼程,彭天强等.智能图像处理技术.北京:电子工业出版社.2004,150
[25]陆宗骐,梁诚.用Sobel算子细化边缘.中国图象图形学报,2000,5(6):516-520
[26]侯仕杰,景晓军.基于边缘检测的改进型图像混合滤波的研究与实现:[硕士学位论文].北京:北京邮电大学,2008
[27]杨道普,马秋禾等.边缘检测Prewitt算子的改进算法.测绘科学, Vol.33 Suppl,Oct.,2008
[28]马宪民,蒋勇.煤矸石二值图像的Roberts快速边缘检测法.仪器仪表学报,2005,第26卷,第8期增刊
[29]李雪,苑玮琦.灰度图像边缘检测算法的性能评价:[硕士学位论文].沈阳:沈阳工业大学,2007
[30] John Canny. A Computational Approach to Edge Detection. IEEE Transactions on pattern analysis and machine intelligence.1998,8(6):679-698
[31] Sankar K.PAL,Robert A.KING. Image Enhancement Using Smoothing with Fuzzy Sets. IEEE Transactions on Systems,Man,and Cybernetics, 1981,11(7):494-501
[32]吴薇.一种改进的模糊边缘检测算法.现代电子技术.2002,4(135):46-48
[33]刘金华,唐竞新等.一种改进的模糊边缘检测快速算法.系统仿真学报,2003,15(2):273-274,297
[34] Ming Zhang,Bahadir Gunturk. A New Image Denoising Method Based on The Bilateral Filter.IEEE,2008:929-932
[35] Jinbo Wu, Zhouping Yin etc. The Fast Multilevel Fuzzy Edge Detection of Blurry Images. IEEE Signal Processing Letters,2007,14(5):344-347
[2] Lee J,Haralick R, Shapiro L.Morphologic edge detection.IEEE Transactions on Robotics and Automation,1987,3(2):142-156
[3]甘金来,刘钊.图像边缘检测算法的比较研究:[硕士学位论文].成都:电子科技大学,2005
[4]冈萨雷斯.数字图像处理(第二版),阮秋琦,阮宇智等.北京:电子工业出版社.2003,464
[5]巴本冬,吴晓娟等.一种快速医学图像亚像素边缘检测方法.山东大学学报(工学版),2005,35(5):63-67
[6]孙红光,李勇.基于双正交小波弱目标边缘提取方法.长春理工大学报,2004,27(2):53-55
[7]周术诚,耿国华.基于自适应阈值的三维图像多尺度边缘检测.西北大学学报(自然科学版),2006,36(3):393-397
[8]金晟业,陈圣波等.遥感图像边缘提取微分方法及其应用.遥感技术与应用,2008,23(6):729-734
[9]谭海峰,赵文杰等.基于过渡区域模糊增强的合成孔径雷达图像边缘提取算法.计算机与现代化,2008,10:24-29
[10]李新,谢永智等.基于运动图像序列的啤酒瓶凸性字符快速提取方法.桂林工学院学报,2005,25(1):97-100
[11]王新,黄兆云.基于多结构元素的数学形态学图像边缘检测.计算机工程与应用,2008,44(7):89-112
[12] Yuqian Zhao,Weihua Gui.Edge Detection Based on Multi-Structure Elements Morphology.IEEE,2006:9795-9798
[13]孙达,刘家锋.基于概率密度梯度的边缘检测.计算机学报,2009,32(2),299-307
[14]冯会真,夏哲雷等.基于神经网络的图像边缘检测方法.浙江:中国计量学院学报,2006,17(4):289-291
[15]张东波,王耀南.FCM聚类算法和粗糙集在医疗图像分割中的应用.仪器仪表学报,2006,27(12):1683-1687
[16]刘曙光,刘明远等.基于Canny准则的基数B样条小波边缘检测.信号处理,2001,17(5):418-423
[17]虞伟,张小刚.遗传算法优化的形态学边缘检测算子.微计算子信息,2008,24(9):168-170
[18]梁光明,孙即祥等. Otsu算法在Canny算子中的应用.湖南:国防科技大学学报,2003,25(5):36-39
[19] Junaid I. Siddique. Kenneth E.BarnerWavelet-based multiresolution edge detection utilizing gray lever edge maps. IEEE,1998:550-554
[20]刘力双,张姚等.图像的快速亚像素边缘检测方法.光电子,2005,16(8),993-996
[21]苏红彬,白凤娥.边缘检测方法及其在医学图像处理中的应用研究.光电子技术与信息,2005,18(2):23-27
[22]薛丽霞,王佐成等.基于多维云空间的多光谱遥感影像边缘检测研究.测绘科学,2008,33(1):188-190
[23]王鹏,耿相铭.基于Prewitt算子的快速帧内编码算法.微计算机信息(测控自动化),2006,22(7-1):192-194
[24]李弼程,彭天强等.智能图像处理技术.北京:电子工业出版社.2004,150
[25]陆宗骐,梁诚.用Sobel算子细化边缘.中国图象图形学报,2000,5(6):516-520
[26]侯仕杰,景晓军.基于边缘检测的改进型图像混合滤波的研究与实现:[硕士学位论文].北京:北京邮电大学,2008
[27]杨道普,马秋禾等.边缘检测Prewitt算子的改进算法.测绘科学, Vol.33 Suppl,Oct.,2008
[28]马宪民,蒋勇.煤矸石二值图像的Roberts快速边缘检测法.仪器仪表学报,2005,第26卷,第8期增刊
[29]李雪,苑玮琦.灰度图像边缘检测算法的性能评价:[硕士学位论文].沈阳:沈阳工业大学,2007
[30] John Canny. A Computational Approach to Edge Detection. IEEE Transactions on pattern analysis and machine intelligence.1998,8(6):679-698
[31] Sankar K.PAL,Robert A.KING. Image Enhancement Using Smoothing with Fuzzy Sets. IEEE Transactions on Systems,Man,and Cybernetics, 1981,11(7):494-501
[32]吴薇.一种改进的模糊边缘检测算法.现代电子技术.2002,4(135):46-48
[33]刘金华,唐竞新等.一种改进的模糊边缘检测快速算法.系统仿真学报,2003,15(2):273-274,297
[34] Ming Zhang,Bahadir Gunturk. A New Image Denoising Method Based on The Bilateral Filter.IEEE,2008:929-932
[35] Jinbo Wu, Zhouping Yin etc. The Fast Multilevel Fuzzy Edge Detection of Blurry Images. IEEE Signal Processing Letters,2007,14(5):344-347