A Multi-feature Descriptor for E-Commodity Image Retrieval
首发时间:2011-10-20
Abstract:With the development of the 3G network and Electronic Commerce, more and more people prefer to shop online, especially via the Cell phone equipped with camera. In order to meet the requirements of the consumers to easily retrieve similar commodity images, we propose a Mobile Image-to-Search system. In this system a content-based retrieval algorithm combining color and shape features for electronic commerce has been investigated, which uses the Normalized Fourier Descriptor and a fuzzy-linking method of color histogram based on the HSV color space. In order to reduce the interference of background, we also use the method of improved canny descriptor before extracting the shape features and the approach of ROI (region of interest) weighted in the color histogram. The system has been tested on Corel data set and our own commodity library, a large number of images selected from Taobao, one of the largest Electronic Commerce web in China. The experimental results indicate that our method has a better retrieval performance.
keywords: commodity image retrieval Normalized Fourier Descriptor color and shape features ROI
点击查看论文中文信息
基于多特征的商品图像检索
摘要:随着3G网络和电子商务的发展,越来越多的人喜欢在网上购物,特别是通过智能手机。为了满足消费者的需求,方便地检索到自己所需的商品,此论文提出了移动图像检索系统。在系统中,提出了一个结合颜色和形状特征,使用标准化的傅里叶描述子和基于HSV空间的模糊颜色直方图的图像检索算法。为了减少图像背景的干扰,本论文在提取形状特征前用改进的canny算子提取轮廓,并且采用对感兴趣区域加权的方式提高模糊颜色直方图的准确性。此论文提出的方法在Corel数据库和从淘宝网上下载的图片库上进行了测试。实验结果表明,论文的算法相比其他基于内容的算法具有更好的检索效果。
基金:
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
基于多特征的商品图像检索
评论
全部评论0/1000