基于环境采样的彩色数字相片智能处理研究与实现
|
摘要
人们日常生活离不开各种证件，而证件相片则通常是必不可少的。随着软、硬件条件的不断提高，彩色数字相片逐渐应用于证件制作，我国正推行的第二代居民身份证就采用了真彩色数字相片。同时，数字相片的处理离不开专业图像处理系统如PhotoShop，CoreDraw等。然而数字图像处理系统的操作是一个交互性的、主观性的过程，由于证件相片在规格、色彩标准上的严格要求，以及制证设备色彩还原的不确定性和证照的海量性等特点，使得利用专业图像系统处理证照显得过于专业、繁琐，标准难以把握，效果不佳。本文致力于对数字相片处理智能化的研究，以弥补以上的困难和不足。
本文的主要研究内容和研究成果包括：
1)基于环境采样的图像增强与复原。证照相片既要调整亮度、对比度等以提高图像的可懂度，又要保持颜色的真实性以提高逼真度，本文通过分析Anil K.Jain等人提出的光线补偿方法的不足，提出了一种对环境光及相机参数进行采样，并归化为修正参数对数字相片进行复原与增强的方法，解决了相片处理过程中标准不一、色彩参差不齐的难题，经过系统在上百个相片采集点运行表明效果良好。同时提出了一种理想的环境采样卡设计模型，具有一定的理论意义。
2)基于人像的单连通性提出了一种运用掩模方法提取人像区域并用形态运算进行边缘修整的相片分割方法。适度的对比度增强可以强化相片的立体感，但对比度增大则易出现高光，在证照印刷、压制过程中高光区域往往容易出现“白斑”现象。针对这一问题，本文提出了一种基于Bezier曲线的S形色阶拉伸曲线模型，并实现了算法。经过该算法调整处理的相片色彩层次丰富，立体感强，高光明显减少。
3)实用的智能处理系统离不开实时的相片检测机制，针对特定的相片规格和色彩要求，本文在人脸检测传统方法和先验知识的前提下，提出了一种基于面部椭圆性的脸宽定量测量方法和其它一些具体的相片规格和色彩检测方法，具有一定的创新性。算法经过近十万张相片检测证明具有良好的鲁棒性。
以上研究成果已经全部在《第二代居民身份证图像智能处理系统》中获得应用与验证，并取得良好效果。通过全国数十家同类系统的比较，本系统在可靠性、实用性等方面具有明显优势。课题的研究源之于实践用之于实践，在实际应用中举一反三取得了良好的应用效果，目前《第二代居民身份证图像智能处理系统》已在全国多个地市正式运行，并获得好评。
Certificate is very important for one person, and a photo of the person is necessary in certificate usually. With the development of the hardware and software, digital photo of all colors is used in certificates, for instance The Second ID Card of china, in which all colors photo is used. On the other hand, the software of image processing like PhotoShop\CoreDraw became a fashion. The digital photo of certificate can be processed by use these software, but the specialized skill and the ability of control the specification are necessary both to do the job. As these factors above, there are some difficulties in process the photo of certificates.
The image enhancement and restitute based of surroundings sampling is one of innovations of the paper. The photo processing of certificate have the particularity, one the one hand, we must adjust brightness and contrast to improve comprehensibility of the photo, one the other hand we must keep the colors like skin to improve the fidelity. The paper analyses the shortage of Lighting Compensation method advanced by Anil KJain, bring a kind of method about color emendation base on Surroundings Sampling Chart and a model of complexion base on YCb'Cr' color space which changed from YCbCr by nonlinear color transformation. Besides, paper gives a perfect design of Surroundings Sampling Chart.
Delete the background of the photo belong the category of image partition. As the portrait is a continuous block, the paper gives a new method by use "mask" chart. Adjust contrast within measure can improve the third dimension of photo, but there are "blaze" when the photo is printed and baked. To solve the question, the paper give a model based of Bezier Curve to extend the hue, and realized the arithmetic.
Real time inspect is necessary to improve practicability of intelligent processing system. To improve robust of digital photo inspect, the paper give a filtration method base on density of pixel. Base on complexion model, the paper give an integrated means by demarcating the eyes firstly. The method of face width measure is another innovation of the paper.
Integrating with the development of The System of The Second ID Card Photo Intelligent Processing, the paper analyses and designs the whole structure of the system. Dependability and practicability are emphases of the system, and especially some technology involved is discussed deeply.
本文的主要研究内容和研究成果包括：
1)基于环境采样的图像增强与复原。证照相片既要调整亮度、对比度等以提高图像的可懂度，又要保持颜色的真实性以提高逼真度，本文通过分析Anil K.Jain等人提出的光线补偿方法的不足，提出了一种对环境光及相机参数进行采样，并归化为修正参数对数字相片进行复原与增强的方法，解决了相片处理过程中标准不一、色彩参差不齐的难题，经过系统在上百个相片采集点运行表明效果良好。同时提出了一种理想的环境采样卡设计模型，具有一定的理论意义。
2)基于人像的单连通性提出了一种运用掩模方法提取人像区域并用形态运算进行边缘修整的相片分割方法。适度的对比度增强可以强化相片的立体感，但对比度增大则易出现高光，在证照印刷、压制过程中高光区域往往容易出现“白斑”现象。针对这一问题，本文提出了一种基于Bezier曲线的S形色阶拉伸曲线模型，并实现了算法。经过该算法调整处理的相片色彩层次丰富，立体感强，高光明显减少。
3)实用的智能处理系统离不开实时的相片检测机制，针对特定的相片规格和色彩要求，本文在人脸检测传统方法和先验知识的前提下，提出了一种基于面部椭圆性的脸宽定量测量方法和其它一些具体的相片规格和色彩检测方法，具有一定的创新性。算法经过近十万张相片检测证明具有良好的鲁棒性。
以上研究成果已经全部在《第二代居民身份证图像智能处理系统》中获得应用与验证，并取得良好效果。通过全国数十家同类系统的比较，本系统在可靠性、实用性等方面具有明显优势。课题的研究源之于实践用之于实践，在实际应用中举一反三取得了良好的应用效果，目前《第二代居民身份证图像智能处理系统》已在全国多个地市正式运行，并获得好评。
Certificate is very important for one person, and a photo of the person is necessary in certificate usually. With the development of the hardware and software, digital photo of all colors is used in certificates, for instance The Second ID Card of china, in which all colors photo is used. On the other hand, the software of image processing like PhotoShop\CoreDraw became a fashion. The digital photo of certificate can be processed by use these software, but the specialized skill and the ability of control the specification are necessary both to do the job. As these factors above, there are some difficulties in process the photo of certificates.
The image enhancement and restitute based of surroundings sampling is one of innovations of the paper. The photo processing of certificate have the particularity, one the one hand, we must adjust brightness and contrast to improve comprehensibility of the photo, one the other hand we must keep the colors like skin to improve the fidelity. The paper analyses the shortage of Lighting Compensation method advanced by Anil KJain, bring a kind of method about color emendation base on Surroundings Sampling Chart and a model of complexion base on YCb'Cr' color space which changed from YCbCr by nonlinear color transformation. Besides, paper gives a perfect design of Surroundings Sampling Chart.
Delete the background of the photo belong the category of image partition. As the portrait is a continuous block, the paper gives a new method by use "mask" chart. Adjust contrast within measure can improve the third dimension of photo, but there are "blaze" when the photo is printed and baked. To solve the question, the paper give a model based of Bezier Curve to extend the hue, and realized the arithmetic.
Real time inspect is necessary to improve practicability of intelligent processing system. To improve robust of digital photo inspect, the paper give a filtration method base on density of pixel. Base on complexion model, the paper give an integrated means by demarcating the eyes firstly. The method of face width measure is another innovation of the paper.
Integrating with the development of The System of The Second ID Card Photo Intelligent Processing, the paper analyses and designs the whole structure of the system. Dependability and practicability are emphases of the system, and especially some technology involved is discussed deeply.
引文
[1] 中华人民共和国公安部，居民身份证制证用数字相片技术要求，GA461—2004
[2] 中华人民共和国公安部，居民身份证总体技术要求
[3] Rafael C．Gonzalez(美)，数字图像处理(第二版)
[4] 周广新，赵龙，数字相片智能检测研究与实现，《计算机仿真》，2005．8
[5] 张宏林，蔡锐，Visual C++数字图像模式识别技术及工程实践，人民邮电出版社，2003．02
[6] 何斌，马天予，Visual C++数字图像处理，人民邮电出版社，2002．12
[7] 孙即祥，《模式识别中的特征提取与计算机视觉不变量》，国防工业出版社，2001．09
[8] Donald Heam,M．Pauline Baker,计算机图形学(第二版)，电子工业出版社
[9] 湖南依地科技，居民身份证图像智能处理系统技术报告，2005．03
[10] 湖南依地科技，居民身份证图像智能处理系统使用说明，2005．03
[11] 刘晨晨，数学形态学图像处理算法应用研究，哈尔滨工业大学硕士论文，2003．01
[12] 於文雪，李松毅，罗立民，基于Windows GDI的图形图像处理探讨，计算机工程，2003．02
[13] 闻芳，周杰，张长水，李衍达，基于局部线性映射神经网络和亮度补偿的彩色人脸检测方法，清华大学学报自然科学版，1999年第39卷第7期
[14] 王晖，洪晓枫，基于API的数字图像处理程序设计，计算机工程与应用，2003．11
[15] 胡冠宇，基于肤色的裸体影像侦测之研究，国立成功大学(台湾)，2004．07
[16] 陈锻生，刘政凯，彩色图像人脸高光区域的自动检测与校正方法，软件学报，2003，14(11)
[17] 万罡，人脸图像检测与识别研究及实践，武汉大学硕士论文，2003．04
[18] 管理，数字图像中基于α值的景象提取算法研究，浙江大学硕士论文，2000．06
[19] 何俊德，基于影像与文字特征之网页内容分类方法之研究，朝阳大学硕士论文(台湾)，2004．06
[20] 孙即祥，现代模式识别，国防科技大学出版社，2002
[21] J.S. Duncan, "Knowledge Directed Left Ventricular Boundary Detection in Equilibrium Radionuclide Angiocardiography", IEEE Tran. Med. Imag, v61.6, pp. 325-336, 1987
[22] J. N. Kaput, P. K. Sahoo, A. K. C. Wong, "A New Method for Gray-level Picture Thresholding Using the Entropy of The Histogram", Comput. Vis. Graph Image Process, vol. 29, pp. 273-285, 1985
[23] J. Kittler, J. Illingworth,"Minimum Error Thresholding", Pattern Recognit vol. 19, pp.41-47, 1986.[24] G.B. Coleman and H. C. Andrews, "Image Segmentation Clustering", Proc. IEEE, vol. 5, pp. 773-785, 1979
[25] 1. Aizenberg, N. Aizenberg, J. hiltner, C. Moraga, E. Meyer zu Bexten, "Cellular Neural Networks and Computational Intelligence in Medical Image Processing", Image and Vision Computing, vol. 19, pp. 177-183, 2001
[26] S. Kobashi, N. Kamiura, Y. Hata, F. Miyawaki, "Volume-quantization-based neural Network Approach to 3D MR Angiograp Image Segmen-tation", Image and Vision Computing, vol. 19, pp. 185-193, 2001
[27] S. Wang, W. Zhu and Z. Liang, "Shape Deformation:SVM Regression and Application to Medical Image Segmentation", International Conference on Computer Vision (ICCV), Vancouver, Canada, 2001
[28] F. Spitzer, Markov Random Fields and Gibbs Ensembles, Amer. Math Mon, vol. 78, pp. 142-154, 1971
[29] R. Kindermann, J. L. Snell, Markov Random Fields and their Applications, Providence: American Mathematical Society, 1980
[30] 魏锴，彩色数字图像中人脸自动检测算法的比较与研究，武汉大学硕士学位论文，2003．05
[31] 李国军，彩色数字图像中人脸自动检测算法的比较与研究，浙江工业大学硕士学位论文，2002．12
[32] 闫胜业，基于学习的人脸面部图像快速检测算法研究，北京工业大学硕士学位论文，003．06
[33] 卢丽敏，图像增强模型及算法研究，国防科学技术大学硕士学位论文，2002．11
[34] 李睿，张贵仓，真彩色数字位图的处理，哈尔滨师范大学自然科学学报，Vol．20，No．1 2004
[35] http://course.cug.edu.cn/rs/COURSE/6-1-2-a.htm 遥感技术课程
[36] http://www.daee-d.com/golden/RGB_colour_01.htm 色彩平衡
[37] http://www.china-vision.net/bbs7中国视觉技术论坛
[38] http://www.pris.edu.cn/imgprocess/theory/theory.htm 数字图像处理专业站
[39] http://www.cgan.net/book/books/print/packcolor/link/5-4_1.html 色彩学理论
[40] http://www.scsky.cn/Article/Class13/Class20/jc/200402/989.html 中性灰的理论探讨与
[2] 中华人民共和国公安部，居民身份证总体技术要求
[3] Rafael C．Gonzalez(美)，数字图像处理(第二版)
[4] 周广新，赵龙，数字相片智能检测研究与实现，《计算机仿真》，2005．8
[5] 张宏林，蔡锐，Visual C++数字图像模式识别技术及工程实践，人民邮电出版社，2003．02
[6] 何斌，马天予，Visual C++数字图像处理，人民邮电出版社，2002．12
[7] 孙即祥，《模式识别中的特征提取与计算机视觉不变量》，国防工业出版社，2001．09
[8] Donald Heam,M．Pauline Baker,计算机图形学(第二版)，电子工业出版社
[9] 湖南依地科技，居民身份证图像智能处理系统技术报告，2005．03
[10] 湖南依地科技，居民身份证图像智能处理系统使用说明，2005．03
[11] 刘晨晨，数学形态学图像处理算法应用研究，哈尔滨工业大学硕士论文，2003．01
[12] 於文雪，李松毅，罗立民，基于Windows GDI的图形图像处理探讨，计算机工程，2003．02
[13] 闻芳，周杰，张长水，李衍达，基于局部线性映射神经网络和亮度补偿的彩色人脸检测方法，清华大学学报自然科学版，1999年第39卷第7期
[14] 王晖，洪晓枫，基于API的数字图像处理程序设计，计算机工程与应用，2003．11
[15] 胡冠宇，基于肤色的裸体影像侦测之研究，国立成功大学(台湾)，2004．07
[16] 陈锻生，刘政凯，彩色图像人脸高光区域的自动检测与校正方法，软件学报，2003，14(11)
[17] 万罡，人脸图像检测与识别研究及实践，武汉大学硕士论文，2003．04
[18] 管理，数字图像中基于α值的景象提取算法研究，浙江大学硕士论文，2000．06
[19] 何俊德，基于影像与文字特征之网页内容分类方法之研究，朝阳大学硕士论文(台湾)，2004．06
[20] 孙即祥，现代模式识别，国防科技大学出版社，2002
[21] J.S. Duncan, "Knowledge Directed Left Ventricular Boundary Detection in Equilibrium Radionuclide Angiocardiography", IEEE Tran. Med. Imag, v61.6, pp. 325-336, 1987
[22] J. N. Kaput, P. K. Sahoo, A. K. C. Wong, "A New Method for Gray-level Picture Thresholding Using the Entropy of The Histogram", Comput. Vis. Graph Image Process, vol. 29, pp. 273-285, 1985
[23] J. Kittler, J. Illingworth,"Minimum Error Thresholding", Pattern Recognit vol. 19, pp.41-47, 1986.[24] G.B. Coleman and H. C. Andrews, "Image Segmentation Clustering", Proc. IEEE, vol. 5, pp. 773-785, 1979
[25] 1. Aizenberg, N. Aizenberg, J. hiltner, C. Moraga, E. Meyer zu Bexten, "Cellular Neural Networks and Computational Intelligence in Medical Image Processing", Image and Vision Computing, vol. 19, pp. 177-183, 2001
[26] S. Kobashi, N. Kamiura, Y. Hata, F. Miyawaki, "Volume-quantization-based neural Network Approach to 3D MR Angiograp Image Segmen-tation", Image and Vision Computing, vol. 19, pp. 185-193, 2001
[27] S. Wang, W. Zhu and Z. Liang, "Shape Deformation:SVM Regression and Application to Medical Image Segmentation", International Conference on Computer Vision (ICCV), Vancouver, Canada, 2001
[28] F. Spitzer, Markov Random Fields and Gibbs Ensembles, Amer. Math Mon, vol. 78, pp. 142-154, 1971
[29] R. Kindermann, J. L. Snell, Markov Random Fields and their Applications, Providence: American Mathematical Society, 1980
[30] 魏锴，彩色数字图像中人脸自动检测算法的比较与研究，武汉大学硕士学位论文，2003．05
[31] 李国军，彩色数字图像中人脸自动检测算法的比较与研究，浙江工业大学硕士学位论文，2002．12
[32] 闫胜业，基于学习的人脸面部图像快速检测算法研究，北京工业大学硕士学位论文，003．06
[33] 卢丽敏，图像增强模型及算法研究，国防科学技术大学硕士学位论文，2002．11
[34] 李睿，张贵仓，真彩色数字位图的处理，哈尔滨师范大学自然科学学报，Vol．20，No．1 2004
[35] http://course.cug.edu.cn/rs/COURSE/6-1-2-a.htm 遥感技术课程
[36] http://www.daee-d.com/golden/RGB_colour_01.htm 色彩平衡
[37] http://www.china-vision.net/bbs7中国视觉技术论坛
[38] http://www.pris.edu.cn/imgprocess/theory/theory.htm 数字图像处理专业站
[39] http://www.cgan.net/book/books/print/packcolor/link/5-4_1.html 色彩学理论
[40] http://www.scsky.cn/Article/Class13/Class20/jc/200402/989.html 中性灰的理论探讨与