一种韩国礼服图像的内外廓形提取算法

doi:10.19398/j.att.202103004

摘要/Abstract

摘要：

为了实现韩国传统礼服款式的信息化,针对韩国礼服高腰大摆的A字廓型和服饰用色搭配较相近的特点,对韩国礼服图像采用了内外廓型分别提取的方法。在外轮廓识别中,采用角度θ和长度c交叉的线形结构形态学算法进行外轮廓提取,并通过二元方差分析对双因子的影响结果进行客观评价,结合主观视觉效果,得出最优外轮廓的组合参数θ=10°,c=9;对于对比度较低的内部饰品轮廓识别,设计CLAHE算法进行图像增强,采用K-Means聚类算法进行领饰图像分割,再用矩形法赋值和基于插值的局部放大修复算法进行填充修复,最后用canny算法进行轮廓提取得出内部领饰轮廓图。结果表明:内外组合的轮廓提取算法可达到主观视觉和客观边界长度评价下的最优效果,是适合韩国礼服图像特点的有效算法。

关键词: 韩国礼服, 图像识别, 线形结构形态学, CLAHE, K-Means

Abstract:

In order to achieve the informatization of the styles of traditionalKorean dresses, this paper extracts the inner and outer contours ofKorean dress images respectively, in view of the characteristics ofKorean dresses, that is, A-shaped contour with high waist and large hemline and similar color collocation in dress. During the recognition of outer contour, the outer contour is extracted using the linear structure morphology algorithm of angle θ and length c, and the effect of two factors was evaluated in an objective way through a binary variance analysis. Combined with the subjective visual effect, the optimal combined parameters for outer contour were obtained: θ=10°, c=9. For the recognition of inner contours of accessories with low contrast, a CLAHE algorithm was designed to enhance image. Secondly, K-means clustering algorithm was adopted to segment the collar image, and then rectangular assignment and local zoom repair algorithm based on interpolation were adopted for filling and repairing. Finally, Canny algorithm was used to extract contours and get the inner contour of collar. The experimental results indicate that the extraction method integrating inner and outer contours can achieve the best effect under subjective vision and objective evaluation of boundary length. It is an effective algorithm for the characteristics ofKorean dress images.

Key words: Korean dress, image recognition, linear structure morphology, CLAHE, K-Means

中图分类号:

TS941.2

高婷, 庄梅玲, 石历丽, 刘静, 王敬雪. 一种韩国礼服图像的内外廓形提取算法[J]. 现代纺织技术, 2022, 30(2): 197-207.

GAO Ting, ZHUANG Meiling, SHI Lili, LIU Jing, WANG Jingxue. An algorithm for extracting inner and outer contours of korean dress images[J]. Advanced Textile Technology, 2022, 30(2): 197-207.

图/表 20

图1 轮廓提取算法流程

Fig.1 The process of contour extraction algorithm

图2 韩国礼服实例

Fig.2 Examples ofKorean dress pictures

图3 礼服外轮廓形态学分析

Fig.3 Morphological analysis of the outer contour of Korean dress

表1 线形结构元素双因素实验结果

Tab.1 Double-factor experimental results of linear structural elements

c值	θ值
c值	0°	10°	20°	30°	40°	50°	60°	70°	80°	90°
0	0	0	0	0	0	0	0	0	0	0
1	0	0	0	0	0	0	0	0	0	0
2	110	0	0	0	0	0	0	0	0	244
3	285	225	343	400	464	467	288	358	63	184
4	384	400	288	288	346	175	175	288	184	460
5	551	551	433	615	503	577	508	275	467	460
6	592	558	503	433	592	523	573	389	167	588
7	592	649	649	649	418	523	629	514	514	414
8	788	815	815	896	725	639	807	514	694	635
9	788	1000	999	896	794	986	807	658	658	582

表2 线形结构元素双因素方差分析结果

Tab.2 Results of double-factor analysis of variance for linear structure elements

来源	平方和	自由度	均方离差	F值	临界值	显著性
因素c	7903402.44	r-1=9	878155.8267	99.27	2	显著
因素θ	221238.24	s-1=9	24582.02667	2.7787	2	显著
误差E	716561.5388	(r-1)(s-1)=81	8846.438751
总和T	8841202.219	rs-1=99

图4 三维交叉影响图

Fig.4 Three dimensional cross influence

图5 θ和c对L的单因素影响

Fig.5 Univariate influence of θ andconL

图6 线形结构元素不同角度轮廓提取

Fig.6 Contour extraction of linear structural elements at different angles

图7 线形结构元素的不同长度轮廓提取

Fig.7 Different length contour extractions of linear structural elements

图8 韩国礼服外轮廓不同方法提取结果

Fig.8 Extracted results of different methods for the outer contour of Korean dress pictures

图9 灰度线性插值计算

Fig.9 Gray linear interpolation calculation

图10 子块2的直方图剪切示意

Fig.10 Histogram clipping diagram of sub-block 2

图11 子块2的直方图

Fig.11 Histogram of sub-block 2

图12 CLAHE处理结果

Fig.12 The CLAHE algorithm processes the results

图13 K-Means领饰分割结果

Fig.13 K-means collar ornament segmentation results

图14 区域填充结果

Fig.14 Region filling results

图15 修复算法示意

Fig.15 Schematic diagram of repair algorithm

图16 填充修复结果

Fig.16 Filling and repair results

图17 轮廓提取结果

Fig.17 Contour extraction results

图18 边缘长度L统计结果

Fig.18 Statistical results of edge length

参考文献 16

[1]	季勇, 王革非. 韩服与汉服之关系研究[J]. 丝绸, 2012, 49(11):76-80.
	JI Yong, WANG Gefei. Study on the relationship between Hanfu and Hanbok[J]. Journal of Silk, 2012, 49(11):76-80.
[2]	李东, 万贤福, 汪军. 采用傅里叶描述子和支持向量机的服装款式识别方法[J]. 纺织学报, 2017, 38(5):122-127.
	LI Dong, WAN Xianfu, WANG Jun. Clothing style recognition approachsupport vectorusing Fourier descriptors andmachines[J]. Journal of Textile Research, 2017, 38(5):122-127.
[3]	孙国栋, 徐亮. 精确型傅里叶高度函数描述子的服装款式识别方法[J]. 中国测试, 2019, 45(8):130-134.
	SUN Guodong, XU Liang. Clothing style recognition approach based on the accurate Fourier heightfunctions shape descriptor[J]. Journal of China Test, 2019, 45(8):130-134.
[4]	XIN A L, LI W. A method for garment design drawing extraction from garment image[C]. Advances in Textile Engineering and Materials. Switzerland: Trans Tech Publications Ltd, 2013: 497-500.
[5]	何晓军, 徐爱功, 李玉. 基于模糊相似性的彩色形态学图像处理方法[J]. 计算机应用研究, 2019, 36(1):258-263.
	HE Xiaojun, XU Aigong, LI Yu. Color morphology image processing method based on fuzzy similarity[J]. Journal of Computer Application Research, 2019, 36(1):258-263.
[6]	ZUIDERVELDK. Contrast Limited Adaptive Histogram Equalization[M]. Graphics Gems. Amsterdam: Elsevier, 1994: 474-485.
[7]	王建, 庞彦伟, 基于CLAHE的x射线行李图像增强[J]. 天津大学学报, 2010, 43(3):194-198.
	WANG Jian, PANG Yanwei, X-ray baggage image enhancement based on CLAHE[J]. Journal of Tianjin University, 2010, 43(3):194-198.
[8]	孙冬梅, 陆剑锋, 张善卿. 一种改进CLAHE算法在医学试纸条图像增强中的应用[J]. 中国生物医学工程学报, 2016, 35(4):502-506.
	SUN Dongmei, LU Jianfeng, ZHANG Shanqing. The application of an improved CLAHE algorithm in image enhancement of medical test strip[J]. Journal of Chinese Biomedical Engineering, 2016, 35(4):502-506.
[9]	刘轩, 刘佳宾. 基于对比度受限自适应直方图均衡的乳腺图像增强[J]. 计算机工程与应用, 2008, 44(10):173-175.
	LIU Xuan, LIU Jiabin. Mammary image enhancement based on contrast limited adaptive histogram equalization[J]. Journal of Computer Engineering and Applications, 2008, 44(10):173-175.
[10]	吴焕丽, 崔可旺, 张馨, 等. 基于改进K-means图像分割算法的细叶作物覆盖度提取[J]. 农业机械学报, 2019, 50(1):42-50.
	WU Huanli, CUI Kewang, ZHANG Xin, et al. Accuracy of fine leaf crop coverage by improvedK-means algorithm[J]. Journal of Agricultural Machinery, 2019, 50(1):42-50.
[11]	马仲林. 基于数字形态学的图像边缘检测算法研究[D]. 兰州:兰州交通大学, 2020.
	MA Zhonglin. Research on Image Edge Detection Algorithm Based on Digital Morphology[D]. Lanzhou: Lanzhou Jiaotong University, 2020.
[12]	周勇. 基于超图的自适应数学形态学研究与应用[D]. 西安:西安电子科技大学, 2019.
	ZHOU Yong. The Research and Application of Adaptive Mathematical Morphology Based on Hypergraphs[D]. Xi'an: Xidian University, 2019.
[13]	李帅, 郭力娜, 曹应举, 等. 基于不同结构元素的地物边界提取对比实验研究[J]. 唐山学院学报, 2018, 31(3):14-19.
	LI Shuai, GUO Lina, CAO Yingju, et al. A comparative and experimental researchon object edge extraction based on different structural elements[J]. Journal of Tangshan University, 2018, 31(3):14-19.
[14]	LIU J, ZHUANG M L, AI R Y, et al. Analysis on the application of image processing technology in clothing pattern recognition[J]. Journal of Donghua University(English Edition), 2020, 37(3):224-231.
[15]	陈浩, 裴瑞杰, 汪鑫. 基于ACO和K-means算法相结合的生菜叶片图像分割方法[J]. 现代计算机, 2019,(21):45-49.
	CHEN Hao, PEI Ruijie, WANG Xin. Image segmentation method of lettuce leaf based on ant colony optimization and K-means CIustering[J]. Journal of Modern Computer, 2019,(21):45-49.
[16]	高强, 高敬阳, 赵地. GNNI U-net:基于组归一化与最近邻插值的MRI左心室轮廓精准分割网络[J]. 计算机科学, 2020, 47(8):213-220.
	GAO Qiang, GAO Jingyang, ZHAO Di. GNNI U-net:Precise segmentation neural network of left ventricular contours for MRI images based on group normalization and nearest interpolation[J]. Journal of ComputerScience, 2020, 47(8):213-220.