基于K-Means和高斯混合模型的云肩色彩提取方法对比

Comparative Study of Cloud Shoulder Color Extraction Methods Based on K-Means and Gaussian Mixture Model

为获取传统服饰图像的色彩构成情况,选取极具特色的服饰品云肩,对比K-means和高斯混合模型的色彩提取效果。对云肩原始图像的R,B,G通道进行降噪预处理;将图像BGR色彩空间转换至RGB与HSV空间,采用肘部法确定最佳类簇数目k值;分别借助两种算法对图像进行分割与主色彩聚类,从执行效率、分割效果和提取精准度3方面进行对比,确定适合云肩图像的色彩提取方法。实验结果表明:相比高斯混合模型,K-means算法对云肩图像色彩的聚类提取效果更优。

In order to obtain the color composition of traditional clothing images,a highly featured clothing item with cloudy shoulder was selected,and the color extraction effect of K-means and Gaussian mixture model were compared.The R,B and G channels of the original image of cloud shoulder were pre-processed for noise reduction.Then the BGR color space of the image was converted to RGB and HSV space,and the elbow method was used to determine the optimal cluster number k value.Then the image was segmented and the main color clustering was performed through the K-means and Gaussian mixture model,the extraction effect was compared from the three aspects of execution efficiency,segmentation effect and extraction accuracy.The experimental results show that compared with the Gaussian mixture model,the K-means algorithm has a better effect on the cluster extraction of the main colors of cloud shoulder images.