联合分割和特征匹配的服装图像检索

Clothing Retrieval via Co-segmentation and Feature Matching

针对目前服装图像"以图搜图"方法主要依赖于服装分割及款式等特征提取的结果,导致检索准确率低的问题,提出一种基于联合分割和特征匹配的服装图像检索方法.首先结合辅助数据集对输入的待检服装图像进行联合分割;然后基于分割出的待检图像的服装区域提取其颜色和Bundled特征,并与检索数据集中的图像特征进行相似度计算,以实现特征匹配;最后按照相似度高低顺序输出检索结果.实验结果表明,该方法能准确地分割并检索出与待检服装相似的服装;在数据集中的分割结果较目前存在的已知方法准确率提高近15%,各类服装的检索准确率均有较大的提高.

As the problem of unsatisfied accuracy of the clothing retrieval,a clothing retrieval method was presentedvia co-segmentation and feature matching.Firstly,the query image was segmented with an auxiliary dataset.Then,the color features and the bundled features of clothing regions were extracted after co-segmentation.The similarities of the features between the query image and the ones in retrieval dataset were calculated.At last,the retrieval results were output in descending order according to the similarity.Experimental results show thatthe proposed method achieves a high segmentation accuracy and retrieval accuracy.The segmentation accuracy isimproved approximately15%comparing with the known segmentation methods,and the retrieval accuracy isgreatly improved in every clothing categories.