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。
In order to achieve the informatization of the styles of traditional Korean dresses,this paper extracts the inner and outer contours of Korean dress images respectively,in view of the characteristics of Korean 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 angleθand 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 of Korean dress images.
作者
高婷
庄梅玲
石历丽
刘静
王敬雪
GAO Ting;ZHUANG Meiling;SHI Lili;LIU Jing;WANG Jingxue(College of Textiles&Clothing,Qingdao University,Qingdao 266071,China;Clothing Department,Xi'an Academy of Fine Arts,Xi'an 710065,China)
出处
《现代纺织技术》
北大核心
2022年第2期197-207,共11页
Advanced Textile Technology
基金
青岛大学研究项目(JXGG2019080)
陕西省科技厅项目(2016FP-07)。
