基于特征的二值商标分层检索方法-《计算机技术与发展》

文章信息/Info

Author(s):: CHU Jing-jing; ZHANG Ji-nan; Computer and Information Engineering College of Henan University

摘要:: 针对以往商标图像搜索方法单一，并且缺少用户反馈的缺点，文中提出一种基于形状特征和纹理特征的二值商标图像分层搜索方法。该方法首先利用Hu不变矩提取图像的形状特征；然后利用欧氏距离来测量图像的相似度，进行第一次筛选得到一个数目不定的候选集；接着利用灰度直方图的方法提取候选集中的图像的纹理信息，进行第二次筛选；最后通过用户的反馈进行不断的优化。从而使检索出来的图像更加迅速和准确，更加符合人的视觉感受

Abstract:: The previous binary trademark image retrieval method is not perfect for lack of comprehensive approach and user feedback. This paper presents a hierarchical binary trademark image retrieval method based on shape and texture features. Firstly, this method uses Hu invariant moments shape feature to extract the images and then uses Euclidean distance to measure the image similarity, and filter a number of uncertainties in the candidate set; then extracts gray level histogram of the image the candidate focused on texture information, and get a second filter; Finally, it is optimized by continuous user feedback. This binary trademark images retrieval toothed is more quickly, accurately, and more in line with human perception