针对电子商务网站二维服装虚拟试穿与搭配展示中模特图像配准的需求,提出了一种基于K-均值聚类和扫描标号法相结合的模特眼睛定位方法.首先对模特图像进行Gamma矫正,进而以K-均值聚类为基础分割图像并完成候选区域的边缘检测,最后采用扫描标号法定位模特眼睛.实验结果表明了该方法的鲁棒性与有效性.%Aiming at the needs of registration for model images in two-dimensional virtual try-on and fashion match presentation in E-Commerce web sites, this paper proposes a model eyes location method based on K-means clustering combined with scanning labeling method. First, gamma correction is taken on the model image, then image segmentation is achieved based on K-means clustering. Followed by image segmentation, edge detection of candidate region is fulfilled. At last, scanning labeling method is used to locate model eyes. Experimental results show the robustness and effectiveness of this mothod.
展开▼