|
倒数粗糙熵图像阈值化分割算法
范九伦 雷 博*
(西安邮电大学通信与信息工程学院 西安 710121)
(电子信息现场勘验应用技术公安部重点实验室 西安 710121)
摘 要:基于粗糙集理论的粗糙熵阈值法不需要图像之外的先验信息。粗糙熵阈值法需要解决两个问题,一是图像信息不完整性的度量,二是图像的粒化。该文基于倒数信息熵,提出一种倒数粗糙熵用来度量图像中信息的不完整性。为了更好地对图像进行粒化,采用一种基于均匀性直方图的粒子选取方式。该文提出的倒数粗糙熵表述简洁,计算简单。实验验证了该文方法的有效性。
关键词:图像处理;阈值分割;粗糙熵;倒数粗糙熵;粒化
1 引言
图像分割是把图像分成各具特性的不同区域并提取出感兴趣目标的方法和过程[1]。图像分割是图像分析、理解和计算机视觉中的难点。在众多的图像分割法中,阈值化分割由于其原理简单、易于实现而被广泛使用。图像阈值化分割技术基于图像的直方图,建立对应的阈值选取准则,寻找最佳的分割阈值。阈值化分割技术已被用于医学图像处理[1,2]、红外目标检测[3]、工业无损检测[4]和遥感图像[5]等领域。基于图像信息利用的不同,阈值化技术大致可分为:基于图像统计信息的阈值法[1]、基于图像模糊信息的阈值法[1]、基于图像粗糙信息的阈值法[6]。
图 3 NDT image1分割结果
图 4 NDT image2分割结果
图 5 OTCBVS\库5\irw02\000215分割结果
其中, BO和F O 分别表示原图像中的目标和背景区域(即,理想分割时对应的目标和背景区域),B T和FT分别表示采用分割算法分割后图像中的目标和背景区域。分类误差 ME的 取值范围为[ 0,1]。M E取值越小,表明分割误差越小,分割后图像的效果越接近理想分割。
图 6 OTCBVS\库5\irw06\000225分割结果
SSIM用来比较两幅图像的结构相似性,其计算公式为
其中, µx 和µ y 分别为图像x 和y 的灰度均值,σ x和σy 分别为图像x 和y 的标准差,σ xy 是图像x 和y 的相关系数。 C1和C 2为常数,以避免分母为0。这里C1 =C2 =0.065 。SSIM的取值范围为[ −1,1],值越大分割效果越好。
图3和图4为6种算法对NDT图像的分割结果。由图3可以看出,此时本文的倒数粗糙熵阈值法分割结果最好。本文算法的分割结果接近理想分割图像,而其他5种算法均失效,不能有效检测NDT图像中的目标区域。对于图4所示图像,罗的方法和本文的倒数粗糙熵算法分割效果比较接近于理想分割图像。其他4种算法不能有效提取图像中的目标区域。
图5和图6列出了2幅OTCBVS库中红外图像的分割结果。图5和图6的两幅图像分别取自库5\irw02数据集和库5\irw06数据集。这两个数据集均显示了两个人进入场景到走出场景的整个过程,本文对数据集中的视频序列进行了测试,图5和图6仅展示了这两个视频序列中的一幅图像。图5展示了两个人进入场景的图像,图6展示了两个人接触的图像。红外图像对比度低,目标比较小。对于这2幅图像,此时最大粗糙熵算法、最大模糊熵算法、罗的方法和最大Masi熵算法失效。最大倒数熵算法对于红外小目标图像有较好的分割效果,因此可以检测出目标区域,但将过多背景错分为目标。本文的倒数粗糙熵算法效果最为理想。
表1列出了6种算法对实验中4幅图像的分割阈值比较。表2分别列出了6种算法对4幅图像分割结果的ME值和SSIM值比较。从表2可以看出,对于图4所示的图像,罗的方法ME值最小,SSIM最大,分割效果最接近理想分割结果,本文的算法次之。对于其他3幅图像本文倒数粗糙熵算法的ME值均是最小,SSIM值最大,因此本文算法的分割效果最好。
表 1 6种算法的阈值比较
表 2 6种算法的ME值与SSIM值比较
5 结束语
粗糙熵阈值法是基于图像局部信息的方法,信息利用的程度取决于粒子的大小,因此合理的粒子大小可以有效提取图像中的弱小目标。本文在已有对数粗糙熵和指数粗糙熵的基础上,定义了倒数粗糙熵,进而提出了一种基于最小倒数粗糙熵的图像阈值分割算法。实验表明,倒数粗糙熵阈值法不仅形式简单,而且可以有效分割NDT图像和红外图像。鉴于研究者已经将变精度的粗糙熵用于图像分割,本文的下一步工作是提出变精度的倒数粗糙熵并用于图像分割。
参 考 文 献
[1]SEZGIN M and SANKUR B. Survey over image thresholding techniques and quantitative performance evaluation[J]. Journal of Electronic Imaging, 2004, 13(1):146–165. doi: 10.1117/1.1631315.
[2]OLIVA D, HINOJOSA S, CUEVAS E, et al. Cross entropy based thresholding for magnetic resonance brain images using Crow Search Algorithm[J]. Expert Systems with Applications, 2017, 79: 164–180. doi: 10.1016/j.eswa.2017.02.042.
[3]聂方彦, 李建奇, 张平凤, 等. 复杂图像的Kaniadakis熵阈值分割方法[J]. 激光与红外, 2017, 47(8): 1040–1045. doi:10.3969/j.issn.1001-5078.2017.08.022.NIE Fangyan, LI Jianqi, ZHANG Pingfeng, et al. Threshold segmentation method of complex image based on Kaniadakis entropy[J]. Laser & Infrared, 2017, 47(8):1040–1045. doi: 10.3969/j.issn.1001-5078.2017.08.022.
[4]NG H F. Automatic thresholding for defect detection[J].Pattern Recognition Letters, 2006, 27(14): 1644–1649. doi:10.1016/j.patrec.2006.03.009.
[5]BHANDARI A K, KUMAR A, and SINGH G K. Tsallis entropy based multilevel thresholding for colored satellite image segmentation using evolutionary algorithms[J]. Expert Systems with Applications, 2015, 42(22): 8707–8730. doi:10.1016/j.eswa.2015.07.025.
[6]PAL S K, SHANKAR B U, and MITRA P. Granular computing, rough entropy and object extraction[J]. Pattern Recognition Letters, 2005, 26(16): 2509–2517. doi:10.1016/j.patrec.2005.05.007.
[7]PAWLAK Z. Rough sets[J]. International Journal of Computer & Information Sciences, 1982, 11(5): 341–356.doi: 10.1007/BF01001956.
[8]PAWLAK Z. Rough Sets: Theoretical Aspects of Reasoning about Data[M]. Dordrecht: Springer, 1991: 2−8.
[9]岳晓冬, 苗夺谦, 钟才明. 基于粗糙性度量的彩色图像分割方法[J]. 自动化学报, 2010, 36(6): 807–816.YUE Xiaodong, MIAO Duoqian, and ZHONG Caiming.Roughness measure approach to color image segmentation[J]. Acta Automatica Sinica, 2010, 36(6):807–816.
[10]吴涛. 图像阈值化的自适应粗糙熵方法[J]. 中国图象图形学报,2014, 19(1): 1–10. doi: 10.11834/jig.20140101.WU Tao. Adaptive rough entropy method for image thresholding[J]. Journal of Image and Graphics, 2014, 19(1):1–10. doi: 10.11834/jig.20140101.
[11]姚龙洋, 张清华, 胡帅鹏, 等. 基于近似集与粒子群的粗糙熵图像分割方法[J]. 计算机科学与探索, 2016, 10(5): 699–708.YAO Longyang, ZHANG Qinghua, HU Shuaipeng, et al.Rough entropy for image segmentation based on approximation sets and particle swarm optimization[J].Journal of Frontiers of Computer Science and Technology,2016, 10(5): 699–708.
[12]刘丽华, 周涛, 周乾智. 基于VPRS粗糙熵的图像分割[J]. 计算机工程与应用, 2018, 54(20): 178–183. doi: 10.3778/j.issn.1002-8331.1804-0090.LIU Lihua, ZHOU Tao, and ZHOU Qianzhi. Image segmentation on entropy of variable precision rough entropy[J]. Computer Engineering and Applications, 2018,54(20): 178–183. doi: 10.3778/j.issn.1002-8331.1804-0090.
[13]SARDAR M, MITRA S, and SHANKAR B U. Iris localization using rough entropy and CSA: A soft computing approach[J]. Applied Soft Computing, 2018, 67: 61–69. doi:10.1016/j.asoc.2018.02.047.
[14]HASSANIEN A E, ABRAHAM A, PETERS J F, et al.Rough sets and near sets in medical imaging: A review[J].IEEE Transactions on Information Technology in Biomedicine, 2009, 13(6): 955–968. doi: 10.1109/TITB.2009.2017017.
[15]SEN D and PAL S K. Generalized rough sets, entropy, and image ambiguity measures[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 2009,39(1): 117–128. doi: 10.1109/TSMCB.2008.2005527.
[16]SEN D and PAL S K. Histogram thresholding using beam theory and ambiguity measures[J]. Fundamenta Informaticae, 2007, 75(1/4): 483–504.
[17]MAŁYSZKO D and STEPANIUK J. Adaptive multilevel rough entropy evolutionary thresholding[J]. Information Sciences, 2010, 180(7): 1138–1158. doi: 10.1016/j.ins.2009.11.034.
[18]邓廷权, 盛春冬. 结合变精度粗糙熵和遗传算法的图像阈值分割方法[J]. 控制与决策, 2011, 26(7): 1079–1082.DENG Tingquan and SHENG Chundong. Image threshold segmentation based on entropy of variable precision rough sets and genetic algorithm[J]. Control and Decision, 2011,26(7): 1079–1082.
[19]吴尚智, 佘志用, 张霞, 等. 利用变精度粗糙熵的图像分割算法[J]. 计算机工程与科学, 2018, 40(10): 1837–1843. doi:10.3969/j.issn.1007-130X.2018.10.016.WU Shangzhi, SHE Zhiyong, HANG Xia, et al. An image segmentation algorithm using variable precision rough entropy[J]. Computer Engineering & Science, 2018, 40(10):1837–1843. doi: 10.3969/j.issn.1007-130X.2018.10.016.
[20]PAL N R and PAL S K. Entropic thresholding[J]. Signal Processing, 1989, 16(2): 97–108. doi: 10.1016/0165-1684(89)90090-X.
[21]吴一全, 占必超. 基于混沌粒子群优化的倒数熵阈值选取方法[J]. 信号处理, 2010, 26(7): 1044–1049. doi: 10.3969/j.issn.1003-0530.2010.07.015.WU Yiquan and ZHAN Bichao. Thresholding based on reciprocal entropy and chaotic particle swarm optimization[J]. Signal Processing, 2010, 26(7): 1044–1049.doi: 10.3969/j.issn.1003-0530.2010.07.015.
[22]吴一全, 殷骏, 毕硕本. 最大倒数熵/倒数灰度熵多阈值选取[J].信号处理, 2013, 29(2): 143–151. doi: 10.3969/j.issn.1003-0530.2013.02.001.WU Yiquan, YIN Jun, and BI Shuoben. Multi-threshold selection using maximum reciprocal entropy/reciprocal gray entropy[J]. Journal of Signal Processing, 2013, 29(2):143–151. doi: 10.3969/j.issn.1003-0530.2013.02.001.
[23]CHENG Hengda and SUN Ying. A hierarchical approach to color image segmentation using homogeneity[J]. IEEE Transactions on Image Processing, 2000, 9(12): 2071–2082.doi: 10.1109/83.887975.
[24]罗钧, 杨永松, 侍宝玉. 基于改进的自适应差分演化算法的二维Otsu多阈值图像分割[J]. 电子与信息学报, 2019, 41(8):2017–2024. doi: 10.11999/JEIT180949.LUO Jun, YANG Yongsong, and SHI Baoyu. Multithreshold image segmentation of 2D Otsu based on improved adaptive differential evolution algorithm[J].Journal of Electronics & Information Technology, 2019,41(8): 2017–2024. doi: 10.11999/JEIT180949.
[25]SHUBHAM S and BHANDARI A K. A generalized Masi entropy based efficient multilevel thresholding method for color image segmentation[J]. Multimedia Tools and Applications, 2019, 78(12): 17197–17238. doi: 10.1007/s11042-018-7034-x.
[26]LI Xueqin, ZHAO Zhiwei, and CHENG H S. Fuzzy entropy threshold approach to breast cancer detection[J].Information Sciences - Applications, 1995, 4(1): 49–56. doi:10.1016/1069-0115(94)00019-x.
[27]http://vcipl-okstate.org/pbvs/bench/, 2013.
[28]WANG Zhou, BOVIK A C, SHEIKH H R, et al. Image quality assessment: From error visibility to structural similarity[J]. IEEE Transactions on Image Processing, 2004,13(4): 600–612. doi: 10.1109/TIP.2003.819861.
Image Thresholding Segmentation Method Based on Reciprocal Rough Entropy
FAN Jiulun LEI Bo
(School of Communication and Information Engineering, Xi’an University of Posts & Telecommunications, Xi’an 710121, China)
(Key Laboratory of Electronic Information Application Technology for Scene Investigation,Public Security Ministry, Xi’an 710121, China)
Abstract: Image thresholding methods based on the rough entropy segment the images without prior information except the images. There are two problems to be considered in the rough entropy based thresholding methods, i.e., measuring the incompleteness of knowledge about an image and granulating the image. In this paper, reciprocal rough entropy, a new form of rough entropy, is defined to measure the incompleteness of the image information. In order to granulate the image effectively, a granule size selection method based on the homogeneity histogram is employed. The proposed reciprocal rough entropy is simple in expression and calculation. The experimental results verify the effectiveness of the proposed algorithm.
Key words: Image processing; Thresholding segmentation; Rough entropy; Reciprocal rough entropy;Granulation
中图分类号:TP391.4
文献标识码:A
文章编号:1009-5896(2020)01-0214-08
DOI: 10.11999/JEIT190559
收稿日期:2019-07-25;改回日期:2019-10-25;网络出版:2019-11-13
*通信作者: 雷博 leileibo@xupt.edu.cn
基金项目:国家自然科学基金(61671377, 61571361, 61601362),西安邮电大学西邮新星团队项目(xyt2016-01)
Foundation Items: The National Natural Science Foundation of China(61671377, 61571361, 61601362), The Project of New Star Team of Xi’an University of Posts & Telecommunications(xyt2016-01)
范九伦:男,1964年生,教授,研究方向为模糊集理论、模糊信息处理、模式识别与图像处理、信息安全.
雷 博:女,1981年生,副教授,研究方向为模糊信息处理、粗糙集理论、图像分割.
|
|