传统服饰云肩实物图像主色的智能检测

发布时间：2019-05-24 16:28

【摘要】：为解决获取传统服饰实物图像色彩耗时且缺乏准确性问题,以传统服饰的云肩为例,借助均值漂移(Mean-shift)聚类法,提出了一种检测传统服饰实物图像颜色的方法。运用单反数码相机进行实物图像采集;对所得初始图像的R、G、B 3个颜色通道进行去噪处理;再将图像RGB颜色空间的特征向量转换至CIE L*a*b*颜色空间中,利用大津法阈值原理(自适应阈值算法)分割被测图像中云肩实物与背景;最后采用Mean-shift聚类算法,将被测图像的颜色像素分割为若干有效的集群,同时从这些集群中提取云肩主要色彩。实验结果表明,该算法可较为准确地从云肩图像中提取主色,且当Mean-shift聚类算法的带宽被设定为0.05时,分类颜色结果更为准确。
[Abstract]:In ord to solve that problem of time-consuming and lack of accuracy of the real image of traditional dress, a method for detecting the color of the real image of traditional dress is proposed by means of the mean-shift clustering method. the method comprises the following steps of: carrying out real-time image acquisition by using a single-inverse digital camera; performing de-noising processing on the R, G, and B three color channels of the obtained initial image; and converting the feature vector of the image RGB color space into the CIE L * a * b * color space, In that pap, the physical and the background of the cloud shoulder in the image to be measured are segmented by the method of the threshold of the large-body method (adaptive threshold); finally, the mean-shift clustering algorithm is adopted, and the color pixels of the image to be measured are divided into several effective clusters, and the main colors of the cloud shoulder are extracted from the clusters. The experimental results show that the algorithm can extract the dominant color from the cloud shoulder image more accurately, and when the bandwidth of the Mean-shift clustering algorithm is set to 0.05, the classification color results are more accurate.
【作者单位】：江南大学纺织服装学院;江苏省服饰类非物质文化遗产研究基地;香港理工大学纺织及制衣学系;
【基金】：中央高校基本科研业务费专项资金项目(JUSRP51417B)
【分类号】：TP391.41;TS941.11

【相似文献】