Advanced Textile Technology ›› 2022, Vol. 30 ›› Issue (2): 197-207.DOI: 10.19398/j.att.202103004

• Apparel Engineering • Previous Articles     Next Articles

An algorithm for extracting inner and outer contours of korean dress images

GAO Ting1, ZHUANG Meiling1(), SHI Lili2, LIU Jing1, WANG Jingxue1   

  1. 1. College of Textiles & Clothing, Qingdao University, Qingdao 266071, China
    2. Clothing Department, Xi'an Academy of Fine Arts, Xi'an 710065, China
  • Received:2021-03-03 Online:2022-03-10 Published:2021-07-08
  • Contact: ZHUANG Meiling

一种韩国礼服图像的内外廓形提取算法

高婷1, 庄梅玲1(), 石历丽2, 刘静1, 王敬雪1   

  1. 1.青岛大学纺织服装学院,山东青岛 266071
    2.西安美术学院服装系,西安 710065
  • 通讯作者: 庄梅玲
  • 作者简介:高婷(1996-),女,山东临沂人,硕士研究生,主要从事服装与服饰数字化方面的研究。
  • 基金资助:
    青岛大学研究项目(JXGG2019080);陕西省科技厅项目(2016FP-07)

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

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

CLC Number: