应用均值漂移的印花面料交互式换色方法  

Interactive color replacement method for printed fabrics based on Mean-shift

作　　者：胡群 张宁[1] 潘如如[1] HU Qun;ZHANG Ning;PAN Ruru(Key Laboratory of Eco-Textiles(Jiangnan University), Ministry of Education, Wuxi, Jiangsu 214122, China)

机构地区：[1]生态纺织教育部重点实验室(江南大学),江苏无锡214122

出　　处：《纺织学报》2021年第11期97-102,共6页Journal of Textile Research

基　　金：国家自然科学基金项目(61976105);中国纺织工业联合会应用基础研究项目(J202006)。

摘　　要：针对印花面料设计专业性要求高、面料打样耗时费力、对面料进行换色生产系列产品困难的问题,提出了一种基于纹理平滑与均值漂移(Mean-shift)的面料分色换色方法。采集并剪裁获得印花面料图像,采用相对总变差模型对面料图像进行纹理平滑,将平滑后的图像转换到CIE1976L^(*)a^(*)b^(*)色彩空间,然后利用Mean-shift聚类算法对面料图像进行分割,提取面料图像分色。通过交互式的方法调节分色,建立目标颜色与图像原始分色的映射关系,实现面料的分色换色,并采集不同的印花面料图像进行换色实验。结果表明:换色图像自然真实,面料纹理细节保留完整,可辅助印花面料设计,模拟印花面料的打样效果。In order to enhance the professional design skills,increase the efficiency in pattern proofing and facilitate color replacement for product series in designing and producing printed fabrics,a color separation and color replacement method was proposed for the printed fabric based on texture smoothing and Mean-shift image segmentation.The printed fabrics were captured and cut to obtain the images before the fabric image was smoothed using the relative total variation model.After transferring into CIE1976L^(*)a^(*)b^(*) color space,the image was segmented using Mean-shift clustering to extract the separated colors.The mapping relationship between new color and original color was built for color replacement.Finally,color replacement was achieved by interactively adjusting values of the separated colors.Different printed fabrics were captured for the color replacement experiments,and the results show that fabric images after color replacement are natural and practical,and texture details of the fabrics are reserved completely.This method can assist the design of printed fabrics as well as effects simulation of fabric proofing.

关 键 词：印花面料 图像平滑 图像分割 Mean-shift聚类 分色换色 

分 类 号：TS194.1[轻工技术与工程—纺织化学与染整工程]