多值模板图像匹配关键问题研究
详细信息    本馆镜像全文|  推荐本文 |  |   获取CNKI官网全文
摘要
数字图像处理技术作为一门专门的研究学科出现以来,其应用已经从最初的工业及商业领域扩展到艺术、文化等领域以及人们的日常生活中。图像匹配是图像识别系统中必不可少的重要环节,也是图像处理中最常见和困难的问题之一。本课题主要针对多值模板匹配方法及其相关的问题进行研究,主要包括以下几个方面。
     模板图像分割方面,首先本文提出了对模板多值化的基于灰度的分割方法,将寻找最优阈值过程建模为寻找使分割后图像与源图像相似度最大阈值集合过程,利用模板匹配公式作为相似度评价标准,并提出了与模板匹配等价的基于直方图的分割算法,理论和实验结果表明,新算法有很好的分割效果,再从非线形规划的角度对最佳阈值的选取过程进行了优化,进一步提高了分割速度。然后提出了基于图像边缘特征为模板多值分割方法,提出了边缘偏离程度的概念,将寻找最优阈值过程建模为寻找使分割后图像与源图像边缘偏离程度最小的阈值集合过程。最后本文将提出的割方法推广到对一般图像的分割,并与OTSU方法,最大熵方法进行了比较。
     快速图像匹配算法方面,本文提出了基于多值模板的图像快速匹配算法,即将最佳多阈值分割后的模板图像作为新的模板进行图像匹配,利用差值模板中存在的大量灰度变化相同区域,采用迭代的方法,减少这些区域的计算,从而减小复杂度,实验得到了比较满意的结果,最后提出了匹配差异的概念,对多值模板匹配结果和源模板的匹配结果差异进行衡量。
     最佳阈值个数选取方面,本文提出了模板图像最佳分割阈值个数的衡量标准和选取条件,目的是用较少的阈值数对源模板进行分割,保证快速匹配的低复杂度和匹配精度。然后研究了不同的图像分割方法对阈值个数的影响,以及阈值个数的选取和模板图像本身特性的关系等相关问题。
Since digital image processing has been a special subject, it was applied in several fields from the initial industry and commerce to the art, culture and people's daily life. The image mate are indispensable components in the pattern recognition system, and it was also one of the most difficult and commonly problem. This topic mainly aims is to research the multi-value template matching method and the key question, Main content include following several aspects.
     In template image segmentation. First, this text proposed the segmentation algorithm which based on gradation to segment template, this threshold segmentation algorithm which transform the process of finding optimal threshold to the process of finding maximal likeness degree with optimal threshold between source picture and after segmentation picture. Taking the template matches formula as the standards of the evaluating likeness degree, and designing a segmentation algorithm based on the histogram which equivalent the template matching. The theoretical and experiment results showed that the new method has good dissection effect, and the speed is quick. Then the text put forward the segmentation algorithm which based on edge to segment template, put forward the concept of edge deviation degree . Finally, this text propose the method to promote to the general picture segmentation.
     In fast image matching, this text put forward the fast-matching method which base on the multi-value template, made the template picture which have been segmented as new template to match image, then utilized the same district of D-template and use a iteration method to reduce the calculation. The experiment got the more satisfied result. Finally, put forward the match difference concept and carry on measuring to the difference between the result of multi-value template matching and Source template matching.
     In the optimal number of threshold, this text chose the standards of how to get the number of image optimal threshold, Selected the best number of threshold to segment source template, Guarantees the low complex in fast matching. Then it has research the different segmentation method to the number of optimal threshold and the relationship betweentemplate characteristic and choice of optimal threshold .
引文
[1] Leila M G, Manjunath B S. Registration techniques for multisensor remotely sensed image. PE&PS, 1996, 62(9): 1049~1056.
    [2] Tsai, D. -M, Lin, C. -T. Fast normalized corss correlation for defect detection. Pattern Recognition. Lett, 2003, 24(15): 2625~2631.
    [3] Stytz, M. R. Frider. Three-dimensional Medical Imaging: Algorithms and Computer Systems. ACM Computer Survey, 1991, 23(4): 421~424
    [4] T.Netsch, P. Rosch, A.Van Muiswinkel, and J. Weese. Towords real-time multi-modality 3-D medical image registration. IEEE Eighth International Conference on Computer Vision. ICCV 2001, 9 (1): 718~725.
    [5] L. G, Brown. A survey of image registration techniques. ACM Computing Surveys, 1992: 326~367.
    [6] Huttenlocher D P, Rucklidge W j. Comparings images using the Hausdorff distance uder translation[A]. IEEE Computer Society Conference on Computer Vision and Pattern Recognition [C]. 1992.
    [7] 王红海,张科,李言俊.图像匹配研究进展.计算机工程与应用[J].2006,26(3):729~732.
    [8] 钟胜,桑农,张天序.基于灰度相关的雷达与可见光景像匹配算法[J].红外与激光工程,1999,28(5):22~25.
    [9] Ji Zhou, JiaoYing shi. A Robust Algorithm for Feature Point Matching[J]. Computers & Graphics, 2002, 26 (3): 429~436.
    [10] Huang Zhaohui, Cohen F S. Affine-invariant B-Spline Moments for Curve Matching[J]. IEEE Trans on Image Processing, 1996, 5 (10): 1473.
    [11] 聂恒,赵荣春,康宝生.基于边缘几何特征的图像精确匹配方法[J].计算机辅助与图形学学报.2004,16(12):1668~1673.
    [12] 夏永泉,刘正东,杨静宇.不变矩方法在区域匹配中的应用[J].计算机辅助与图形学学报.2005,17(10):2152~2156.
    [13] 张伟,吴刚,侯晴宇等.基于不变矩特征匹配的目标定位方法的实现[J].光学技术.2005,31(3):441~444.
    [14] 沈海滨,赖女.基于图像中心矩的快速模板匹配方法[J].计算机应用,2004,24(11):116~118.
    [15] 李培华,张田文.主动轮廓线模型(蛇模型)综述[J].软件学报,2000,11(6):751~757.
    [16] 张桂林,徐捷.频域相关技术在图像匹配中的应用[J].模式识别与人工智能,1997,10(1):87~92.
    [17] 李强,张钹.一种基于图像灰度的快速匹配算法[J].软件学报,2006,17(2):216~222.
    [18] Mark Holden, Derek L G Hill, Erika R E Denton. Voxel similarity measures for 3-D serial MR brain image registration. IEEE Teansaction on Medical Image. 2000, 19(2): 94~102.
    [19] 宋毅,催远平,居鹤华.一种图像匹配中SSDA和NCC酸法的改进[J].计算机工程与应用,2006,17(12):42~44.
    [20] Jiang Xiaoyu, Huang Yingqing. Multi-resolution template match using wavelet transform[J]. Journal of Image and Graphics, 2000, 5A (4): 304~308.
    [21] Sun Yuan, Zhou Ganghui, Zhao Lichu Fast template matching algorithm based on the projection[J]. Journal of Shanghai Jiaotong University, 2000, 34 (5): 702~704(in Chinese).
    [22] Shen Tingzhi, Fang Ziwen. Digital image processing and pattern recognition[M]. Beijing, 1998.
    [23] 沈庭芝,方子文.数字图像处理及模式识别[M].北京:北京理工大学出版社,1998.
    [24] 熊国清,于起峰.用于实时跟踪的快速匹配算法[J].计算机辅助设计与图形学学报,2002,14(1):41~43.
    [25] 严柏军,郑链,王克勇.基于不变矩特征匹配的快速目标检测算法[J].红外技术,2001,23(6):.
    [26] Shen Tingzhi, Fang Ziwen. Digital image processing and pattern recognition[M]. Beijing, 1998.
    [27] 沈庭芝,方子文.数字图像处理及模式识别[M].北京:北京理工大学出版社,1998.
    [28] Otsu.N. A threshold selection method from grey-level histograms[J]. IEEE Trans.System,Man and Cybernetic, 1979, 9 (1): 62~66.
    [29] 郭海涛,田坦,王连玉等.利用二维属性直方图的最大熵的图像分割方法[J].光学学报,2006,26(4):506~509.
    [30] 傅建平,廖振强,张培林等.基于2D熵阈值的铁谱磨粒图像分割方法[J].计算机工程与应用,2005,18(2):204~206.
    [31] Kapur J N, P K Sahoo, A K C Wong. A new method for gray-level picture thresholding using the entropy of the histogram[J]. Computer Vision, Graphics and Image Processing, 1985, 29 (3): 273~285.
    [32] 黄文芝,王以治.基于X~2分布的聚类图像分割[J].中国图像图形学报,2004,9(2):164~167.
    [33] 匡继昌.实分析与泛函分析[M].北京:高等教育出版社:116-117,2002.
    [34] 希梅尔布劳.实用非线性规划[M].张义桑译,北京:科学出版社,1981.
    [35] 魏伟波,芮筱亭.图像边缘检测方法研究[J].计算机工程与应用,2006,30(3):88~91.
    [36] 王成儒,倪永婧.基于精度准则的图像分割算法评价[J].微计算机信息,2006,22(5-1):248~230.
    [37] 侯格贤,毕驾彦.图像分割质量评价方法研究[J].中国图像图形学报,2000,5(1):39~43.
    [38] 孙远,周刚慧,赵立初.灰度图像匹配的快速算法[J].上海交通大学学报,2000,34(5):702~704.
    [39] 韩瑞峰,张永奎.一种改进的实数编码遗传算[J].计算机工程与应用,2002,38(3):78~80.
    [40] 李强,张钹.一种基于图像灰度的快速匹配算法[J].软件学报,2006,17(2):217~222.
    [41] 魏宝刚,鲁东明,潘云鹤等.多颜色空间上的交互式图像分割[J].计算机学报,2001,24(7):770~775.
    [42] 马义德,李廉,戴若兰.基于细胞逻辑形态特征图像分割新算法[J].电子科技大学学报,2002,31(1):84~87.

© 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