基于边缘和角点的图像特征提取方法的研究及实现
|
摘要
随着科学技术的进步,人类社会已经迈入了一个全新的数字化时代。图像作为信息的一种重要载体,对其进行有效的研究和表示,在数字信息处理中有着非常重要的意义。图像特征作为图像中可用做标注的属性,常常成为数字图像研究领域的热点和难点,正确提取图像的特征是图像分割、图像理解、模式识别和计算机视觉等领域的研究基础和关键前提。在诸多图像特征中,尤以图像的边缘和角点最为重要,因为它们能比较完整地刻画出图像的特点。图像的边缘可定义为图像的灰度、纹理、颜色等局部特征不连续性的区域,角点属于边缘中最特殊的一类点,它是尖锐边缘的端点,角点还常常被定义为图像边界上曲率足够高或曲率变化足够明显的点。图像的边缘检测和角点检测是图像特征提取领域最为重要的研究内容,也是后续高级图像处理的基础。
本文首先细致比较了各类经典边缘检测算法以及SUSAN角点检测算法中的优劣点,发现传统算法中存在缺乏滤波机制或者滤波机制较弱,缺少办法对滤波后的模糊化图像特征进行加强,没用使用多尺度的方式来降低图像特征的多重响应和降低由噪声干扰引起的图像特征误检。针对以上不足,本文提出了一系列新的解决方案,主要的工作包括以下三个方面:
1)针对传统算法中对图像的滤波效果较差的情况,提出了一组共三个保留图像细节的改进滤波算法(分别为CS-LAMF,ASF-M,CS-GF),来分别处理椒盐噪声、高斯噪声和混合噪声并尽可能地降低图像特征模糊化现象。
2)引入数学形态学的相关理论,利用其运算对滤波后的图像进行了特征加强处理,并结合新提出的一组滤波算法提出一种基于Canny的噪声类型自适应的边缘检测新算法。
3)借鉴了Harris角点检测算法中的设计思想,融合了本文新提出的伪角点模板匹配检测法、自适应计算USAN阈值和初始角点响应值上限/下限抑制这三种新机制给出一种改进的基于SUSAN的综合改进的自适应角点检测算法。
通过本文最后的对比实验可以看出,新提出的一组滤波机制在有效改进滤波效果的同时很好地维持了图像特征的细节;基于Canny的噪声类型自适应的边缘检测算法则在多类噪声环境下有着远超传统算法的检测效果;基于SUSAN的综合改进的自适应角点检测算法既能提供较好的检测效果也保持了较好的实时性能。
With the development of social science and technology, human society has entered into a brand-new digital era. As an important carrier of information, it is significant to do efficient research and expression with image. Image character can be used as the property of notes in the image, however, at the same time, this becomes a hot and difficult area in the digital image research. Correctly extracting image property is the research basis and key premise of image segmentation, image understanding, pattern recognition and computer vision. Within all of the image characteristics, image edge and corner are of greatest importance, because they can relatively depict the characteristics of the image thoroughly. Image edge can be defined as the discontinued area of partial character of image grayscale, textures and color, while corner is the most peculiar type of edge corner. It is the terminal of sharp edge and usually it can be defined as the corner where curvature is high enough or curvature that changes often enough as well. Image edge detection and corner detection are the crucial research content in the image character extract area and the basis of subsequent image processing.
Firstly, this paper thoroughly compares the main classic edge detection algorithm and analyses the advantage and disadvantages in the traditional SUSAN corner algorithm. Through the comparison, it is found that there are lacks of noise filter system or the filter system is relatively weak and a lack in the methods to strength the fuzzy image character, and there are no any multi-scale methods to reduce the multi-response from image characteristics or incorrect image characteristics detection which is caused by noise interference neither. According to all weakness raised above, this paper provides a series of new solutions, the main content includes three aspects shown as below.
1) The paper counters the situation of the bad effect of noise filter with a set algorithm composed of three improved filter algorithms (CS-LAMF,ASF-M,CS-GF) which reserve image details to deal with pepper noise, Gaussian noise and mixed noise respectively, meanwhile it can reduce the fuzziness of the image characteristics.
2) It also introduces some related morphological theory in mathematics to strengthen the character property after wave filtering and propose a brand-new edge detection algorithm based on Canny which is also self-adaptive to noise type.
3) It finally raises a new self-adaptive improved corner detection algorithm based on SUSAN which is combined the idea from Harris algorithm's design with using default template to match and eliminate fake corner, making SUSAN threshold self-adaptive and setting upper limit/low limit to the original corner response value.
From the comparative lab data, we can find the new filter system can finely maintain details of the image property while improve the filter effects at the same time, the new edge detection algorithm has detected better results than traditional algorithm in the condition of multi type noises, the proposed corner detection algorithm also can provide better detection results while reserve a good real time performance as well.
本文首先细致比较了各类经典边缘检测算法以及SUSAN角点检测算法中的优劣点,发现传统算法中存在缺乏滤波机制或者滤波机制较弱,缺少办法对滤波后的模糊化图像特征进行加强,没用使用多尺度的方式来降低图像特征的多重响应和降低由噪声干扰引起的图像特征误检。针对以上不足,本文提出了一系列新的解决方案,主要的工作包括以下三个方面:
1)针对传统算法中对图像的滤波效果较差的情况,提出了一组共三个保留图像细节的改进滤波算法(分别为CS-LAMF,ASF-M,CS-GF),来分别处理椒盐噪声、高斯噪声和混合噪声并尽可能地降低图像特征模糊化现象。
2)引入数学形态学的相关理论,利用其运算对滤波后的图像进行了特征加强处理,并结合新提出的一组滤波算法提出一种基于Canny的噪声类型自适应的边缘检测新算法。
3)借鉴了Harris角点检测算法中的设计思想,融合了本文新提出的伪角点模板匹配检测法、自适应计算USAN阈值和初始角点响应值上限/下限抑制这三种新机制给出一种改进的基于SUSAN的综合改进的自适应角点检测算法。
通过本文最后的对比实验可以看出,新提出的一组滤波机制在有效改进滤波效果的同时很好地维持了图像特征的细节;基于Canny的噪声类型自适应的边缘检测算法则在多类噪声环境下有着远超传统算法的检测效果;基于SUSAN的综合改进的自适应角点检测算法既能提供较好的检测效果也保持了较好的实时性能。
With the development of social science and technology, human society has entered into a brand-new digital era. As an important carrier of information, it is significant to do efficient research and expression with image. Image character can be used as the property of notes in the image, however, at the same time, this becomes a hot and difficult area in the digital image research. Correctly extracting image property is the research basis and key premise of image segmentation, image understanding, pattern recognition and computer vision. Within all of the image characteristics, image edge and corner are of greatest importance, because they can relatively depict the characteristics of the image thoroughly. Image edge can be defined as the discontinued area of partial character of image grayscale, textures and color, while corner is the most peculiar type of edge corner. It is the terminal of sharp edge and usually it can be defined as the corner where curvature is high enough or curvature that changes often enough as well. Image edge detection and corner detection are the crucial research content in the image character extract area and the basis of subsequent image processing.
Firstly, this paper thoroughly compares the main classic edge detection algorithm and analyses the advantage and disadvantages in the traditional SUSAN corner algorithm. Through the comparison, it is found that there are lacks of noise filter system or the filter system is relatively weak and a lack in the methods to strength the fuzzy image character, and there are no any multi-scale methods to reduce the multi-response from image characteristics or incorrect image characteristics detection which is caused by noise interference neither. According to all weakness raised above, this paper provides a series of new solutions, the main content includes three aspects shown as below.
1) The paper counters the situation of the bad effect of noise filter with a set algorithm composed of three improved filter algorithms (CS-LAMF,ASF-M,CS-GF) which reserve image details to deal with pepper noise, Gaussian noise and mixed noise respectively, meanwhile it can reduce the fuzziness of the image characteristics.
2) It also introduces some related morphological theory in mathematics to strengthen the character property after wave filtering and propose a brand-new edge detection algorithm based on Canny which is also self-adaptive to noise type.
3) It finally raises a new self-adaptive improved corner detection algorithm based on SUSAN which is combined the idea from Harris algorithm's design with using default template to match and eliminate fake corner, making SUSAN threshold self-adaptive and setting upper limit/low limit to the original corner response value.
From the comparative lab data, we can find the new filter system can finely maintain details of the image property while improve the filter effects at the same time, the new edge detection algorithm has detected better results than traditional algorithm in the condition of multi type noises, the proposed corner detection algorithm also can provide better detection results while reserve a good real time performance as well.
引文
[1]K.R.Castleman.,Digital Image Processing[M],Tsinghua University Publishing House,Beijing,1998
[2]冈萨雷斯,数字图像处理[M],电子工业出版社,2004
[3]王建勇,周晓光,廖启征,一种基于中值-模糊技术的混合噪声滤波器[J],电子与信息学报,vol.28,No.5,2006,pp:901-904
[4]卢京晶、方中华、孙胜利,一种自适应的加权均值滤波器[J],传感技术学报,vol.18,No.4,2005,pp:880-882
[5]高浩军、杜宇人,中值滤波在图像处理中的应用[J],电子工程师,Vol.30,No.8,2003
[6]Sicuranza G.L.,Theory and Realization of Nonlinear Digital Filters[J],ISCAS-84,1981,pp:242-245
[7]T.Poggio,H.Voorhees and A.Yuille,A Regularized Solution to Edge Detection,Tech.Rep.MA,Rep.AIM-833,MIT Articial Intell.Lab.,May 1985
[8]Roberts L.G..,Machine Perception of Three-dimension Solids,Optical and Electro-optical information processing,Tippet,J.T.(ed.),MIT Press,Cambridge,Mass,1965
[9]Prewitt J.M.S,Object Enhancement and Extraction[J],Picture Processing and psychopictorics.,1970,pp:75-149
[10]L.Sobel,Camera Models and Machine Perception[Dissertation],PhD theses,Stanford University,Standford,CA,1970
[11]Canny J.,A Computational Approach for Edge Detection[J],IEEE Trans.Pattern Anal.Machine Intell.,1986,8(6):679-698
[12]Fram J.R.,Deutsch E.S.,On the Quantitative Evaluation of Edge Detection Schemes and Their Comparison with Human Performance[J],IEEE Trans.Computers,1975,C-24(6):616-628
[13]S.Kurakake,R.Nevatia,Description and Tracking of Moving Articulated Objects,In 11th Intl.Conf.on Pattern Recognition,Hague,Netherlands,1992,volume 1:491-495
[14]Moravec H.P.,Towards Automatic Visual Obstacle Avoidance,Proc.5th International Joint Conference on Artificial Intelligence,1977,pp:584
[15]Copper J,Svetha,Kitchen L,Early Jump-out Corner Detectors[J],IEEE Trans on PAMI,1993,15:823-828
[16]Hsin.T,Hu W.C,A Rotationally Invariant Two Phase Scheme for Corner Detection[J],Pattern Recognition,1996,28:819-828
[17]In R.S,Chu C.H,Hsueh Y.C,A Modified Morphological Corner Detector[J],Pattern Recognition Letters,1998,129:279-286
[18]Ovalev V.A,Petrou M.,Bondar.Y.S,Using Orientation Tokens for Object Recognition[J],Pattern Recognition,1998,119:1125-1132
[19]Gennery D.B,Object Detection and Measurement using Stereo Vision,Proc ARPA Image Understanding Workshop,College Park,MD,1980,pp:161-167
[20]Beaudet P.R.,Rotational Invariant Image Operators,Proc.4~(th) International Conference on Pattern Recognition,1978,pp:579-583
[21]Harris C.,Stephens M.,A Combined Corner and Edge Detector,Proc 4~(th) Alvey Vision Conference,1988,pp:189-192
[22]Smith S,Brady M.,A new approach to low level image processing[J],International Journal of Computer Vision,1997,23(1):45-78
[23]Haralick.,Statistical and Structural Approaches to Texture[J],Proc.IEEE,1979,67(5):786-804
[24]Levine M.,Vision in Man and Machine[M],McGraw-Hill,1985
[25]Weszka J.,Deya C.,Rosenfeld A.,A Comparative Study of Texture Measures for Terrain Classification[J],IEEE Trans.System,Man and Cybernetics,1976(6):269-285
[26]Cross G.,Jain A.,Markov Random Field Texturemodels[J],IEEE Trans.Pattern Analysis and Machine Intelligence,1983,5(1):25-39
[27]Pentland A.,Fractal-based Description of Natural Scenes,IEEE Trans.Pattern Analysis and Machine Intelligence,1984,6(6):661-674
[28]Strzelechi M.,Materka A.,Markov Random Fields as Models of Textured Biomedical Images,Proc.20~(th) National Conf.Circuit Theory and Electronic Networks KTOiUE'97,Kolobrzeg,Poland,1997,pp:493-498
[29]Rosenfeld A.,Weszka J.,Picture Recognition in Digital Pattern Recognition,K.Fu(Ed.),Springer-Verlag,1980,pp:135-166
[30]Bovik A.,Clark M.,Giesler W.,Multichannel Texture Analysis using Localised Spatial Filters[J],IEEE Trans.Pattern Analysis and Machine Intelligence,1990,12:55-73
[31]Mallat S.,Multifrequency Channel Decomposition of Images and Wavelet Models[J],IEEE Trans.Acoustic,Speech and Signal Processing,1989,37(12):2091-2110
[32]Lu C.,Chung P.,Chen C.,Unsupervised Texture Segmentation via Wavelet Transform,Pattern Recognition,1997,30(5):729-742
[33]G.G.Lorentz著,谢庭藩等泽.Approximation Of Functions《函数逼近论》[M],上海科学技术出版社,1981
[34]R.Kirsch,Computer Determination of the Constituent Structure of Biological Images[J],Computers and Biomedical Research,Vol.4,1971
[35]D.C.Marr,E.Hildreth,Theory of Edge Detection[M],Proc.Roy.Soc.,London,Vol.B275,pp.187-217,1980
[36]John Canny,Finding Edges and Lines in Images,MIT Artif.Intell.Lab.,Cambridge,MA,Tech.Rep.AI-TR-720,1983
[37]I.E.Abdou,W.K.Pratt,Quantitative Design and Evalution of Enhancement and Thrsholding Edge Detectors[J],Proc.IEEE,May 1979,67(5):753-763
[38]W.K.Pratt,Digital Image Processing[M],Wiley-Interscience,New York,1978
[39]孙洪雨,基于数学形态学的图像增强与边缘检测方法的研究,大连理工大学,大连,2005
[40]Ioannis Matalas,Ralph Benjamin,Richard Kitney,An Edge Detection Technique Using the Facet Model and Parameterized Relaxation Labeling[J],IEEE Trans.on Pattern Analysis and Machine Intelligence,1997,19(4):328-341
[41]陈乐,吕文阁,丁少华,角点检测技术研究进展[J],自动化技术与应用,2005,22(4):258-260
[42]Petez M.M.,Dennis T.J.,An Adaptive Implementation of the SUSAN Method for Image Edge and Feature Detection[A],Proceedings of International Conference on Image Processing,Santa Barbara,1997,2:394-397
[43]席志红,刘利彬,许新利,SUSAN角点检测算法改进[J],应用技术,2006,33(9):29-31
[44]何斌,马天予,王运坚,Visual C++数字图像处理[M],北京,人们邮电出版社:2001
[45]王郑耀,数字图像的边缘检测[D],西安:两安交通大学,2003
[46]刘清,林土胜,基于数学形态学的图像边缘检测算法[J],华南理工大学学报(自然科学版),2008,36(9):113-116
[47]曾萍萍,吴建华,多尺度顺序形态学灰度图像边缘检测方法[A],第十二届全国图象图形学学术会议论文集,北京:中国图象图形学学会,2005:257-262
[48]景晓军,李剑峰,熊玉庆,静止图像的一种自适应平滑滤波算法[J],通信学报,2002,23(101:6-14
[49]Trajkovic M.,Hedley M.,Fast corner detection[J],Image and Vision Computing,1998,18:75-87
[50]庄志国,孙惠军等,基于角点检测的图像匹配算法及其在图像拼接中的应用[J],厦门大学学报(自然科学版),2007,46(4):501-505
[2]冈萨雷斯,数字图像处理[M],电子工业出版社,2004
[3]王建勇,周晓光,廖启征,一种基于中值-模糊技术的混合噪声滤波器[J],电子与信息学报,vol.28,No.5,2006,pp:901-904
[4]卢京晶、方中华、孙胜利,一种自适应的加权均值滤波器[J],传感技术学报,vol.18,No.4,2005,pp:880-882
[5]高浩军、杜宇人,中值滤波在图像处理中的应用[J],电子工程师,Vol.30,No.8,2003
[6]Sicuranza G.L.,Theory and Realization of Nonlinear Digital Filters[J],ISCAS-84,1981,pp:242-245
[7]T.Poggio,H.Voorhees and A.Yuille,A Regularized Solution to Edge Detection,Tech.Rep.MA,Rep.AIM-833,MIT Articial Intell.Lab.,May 1985
[8]Roberts L.G..,Machine Perception of Three-dimension Solids,Optical and Electro-optical information processing,Tippet,J.T.(ed.),MIT Press,Cambridge,Mass,1965
[9]Prewitt J.M.S,Object Enhancement and Extraction[J],Picture Processing and psychopictorics.,1970,pp:75-149
[10]L.Sobel,Camera Models and Machine Perception[Dissertation],PhD theses,Stanford University,Standford,CA,1970
[11]Canny J.,A Computational Approach for Edge Detection[J],IEEE Trans.Pattern Anal.Machine Intell.,1986,8(6):679-698
[12]Fram J.R.,Deutsch E.S.,On the Quantitative Evaluation of Edge Detection Schemes and Their Comparison with Human Performance[J],IEEE Trans.Computers,1975,C-24(6):616-628
[13]S.Kurakake,R.Nevatia,Description and Tracking of Moving Articulated Objects,In 11th Intl.Conf.on Pattern Recognition,Hague,Netherlands,1992,volume 1:491-495
[14]Moravec H.P.,Towards Automatic Visual Obstacle Avoidance,Proc.5th International Joint Conference on Artificial Intelligence,1977,pp:584
[15]Copper J,Svetha,Kitchen L,Early Jump-out Corner Detectors[J],IEEE Trans on PAMI,1993,15:823-828
[16]Hsin.T,Hu W.C,A Rotationally Invariant Two Phase Scheme for Corner Detection[J],Pattern Recognition,1996,28:819-828
[17]In R.S,Chu C.H,Hsueh Y.C,A Modified Morphological Corner Detector[J],Pattern Recognition Letters,1998,129:279-286
[18]Ovalev V.A,Petrou M.,Bondar.Y.S,Using Orientation Tokens for Object Recognition[J],Pattern Recognition,1998,119:1125-1132
[19]Gennery D.B,Object Detection and Measurement using Stereo Vision,Proc ARPA Image Understanding Workshop,College Park,MD,1980,pp:161-167
[20]Beaudet P.R.,Rotational Invariant Image Operators,Proc.4~(th) International Conference on Pattern Recognition,1978,pp:579-583
[21]Harris C.,Stephens M.,A Combined Corner and Edge Detector,Proc 4~(th) Alvey Vision Conference,1988,pp:189-192
[22]Smith S,Brady M.,A new approach to low level image processing[J],International Journal of Computer Vision,1997,23(1):45-78
[23]Haralick.,Statistical and Structural Approaches to Texture[J],Proc.IEEE,1979,67(5):786-804
[24]Levine M.,Vision in Man and Machine[M],McGraw-Hill,1985
[25]Weszka J.,Deya C.,Rosenfeld A.,A Comparative Study of Texture Measures for Terrain Classification[J],IEEE Trans.System,Man and Cybernetics,1976(6):269-285
[26]Cross G.,Jain A.,Markov Random Field Texturemodels[J],IEEE Trans.Pattern Analysis and Machine Intelligence,1983,5(1):25-39
[27]Pentland A.,Fractal-based Description of Natural Scenes,IEEE Trans.Pattern Analysis and Machine Intelligence,1984,6(6):661-674
[28]Strzelechi M.,Materka A.,Markov Random Fields as Models of Textured Biomedical Images,Proc.20~(th) National Conf.Circuit Theory and Electronic Networks KTOiUE'97,Kolobrzeg,Poland,1997,pp:493-498
[29]Rosenfeld A.,Weszka J.,Picture Recognition in Digital Pattern Recognition,K.Fu(Ed.),Springer-Verlag,1980,pp:135-166
[30]Bovik A.,Clark M.,Giesler W.,Multichannel Texture Analysis using Localised Spatial Filters[J],IEEE Trans.Pattern Analysis and Machine Intelligence,1990,12:55-73
[31]Mallat S.,Multifrequency Channel Decomposition of Images and Wavelet Models[J],IEEE Trans.Acoustic,Speech and Signal Processing,1989,37(12):2091-2110
[32]Lu C.,Chung P.,Chen C.,Unsupervised Texture Segmentation via Wavelet Transform,Pattern Recognition,1997,30(5):729-742
[33]G.G.Lorentz著,谢庭藩等泽.Approximation Of Functions《函数逼近论》[M],上海科学技术出版社,1981
[34]R.Kirsch,Computer Determination of the Constituent Structure of Biological Images[J],Computers and Biomedical Research,Vol.4,1971
[35]D.C.Marr,E.Hildreth,Theory of Edge Detection[M],Proc.Roy.Soc.,London,Vol.B275,pp.187-217,1980
[36]John Canny,Finding Edges and Lines in Images,MIT Artif.Intell.Lab.,Cambridge,MA,Tech.Rep.AI-TR-720,1983
[37]I.E.Abdou,W.K.Pratt,Quantitative Design and Evalution of Enhancement and Thrsholding Edge Detectors[J],Proc.IEEE,May 1979,67(5):753-763
[38]W.K.Pratt,Digital Image Processing[M],Wiley-Interscience,New York,1978
[39]孙洪雨,基于数学形态学的图像增强与边缘检测方法的研究,大连理工大学,大连,2005
[40]Ioannis Matalas,Ralph Benjamin,Richard Kitney,An Edge Detection Technique Using the Facet Model and Parameterized Relaxation Labeling[J],IEEE Trans.on Pattern Analysis and Machine Intelligence,1997,19(4):328-341
[41]陈乐,吕文阁,丁少华,角点检测技术研究进展[J],自动化技术与应用,2005,22(4):258-260
[42]Petez M.M.,Dennis T.J.,An Adaptive Implementation of the SUSAN Method for Image Edge and Feature Detection[A],Proceedings of International Conference on Image Processing,Santa Barbara,1997,2:394-397
[43]席志红,刘利彬,许新利,SUSAN角点检测算法改进[J],应用技术,2006,33(9):29-31
[44]何斌,马天予,王运坚,Visual C++数字图像处理[M],北京,人们邮电出版社:2001
[45]王郑耀,数字图像的边缘检测[D],西安:两安交通大学,2003
[46]刘清,林土胜,基于数学形态学的图像边缘检测算法[J],华南理工大学学报(自然科学版),2008,36(9):113-116
[47]曾萍萍,吴建华,多尺度顺序形态学灰度图像边缘检测方法[A],第十二届全国图象图形学学术会议论文集,北京:中国图象图形学学会,2005:257-262
[48]景晓军,李剑峰,熊玉庆,静止图像的一种自适应平滑滤波算法[J],通信学报,2002,23(101:6-14
[49]Trajkovic M.,Hedley M.,Fast corner detection[J],Image and Vision Computing,1998,18:75-87
[50]庄志国,孙惠军等,基于角点检测的图像匹配算法及其在图像拼接中的应用[J],厦门大学学报(自然科学版),2007,46(4):501-505