您的位置:首页 > 论文页面
基于区域映射与全概率模型的多示例图像检索方法
发表时间:2017-08-31 浏览量:1766 下载量:424
全部作者: | 吴建国,李辉 |
作者单位: | 安徽大学计算机科学与技术学院 |
摘 要: | 针对图像的语义检索问题,提出基于区域映射与全概率模型(total probability model,TPM)的多示例检索方法。该方法将图像看作多示例包、图像中的区域看作示例,首先对图像进行简单线性迭代聚类(simple linear iterative clustering,SLIC)超像素分割,再寻找图像中的关键点,通过图像区域映射取出关键点所在的超像素区域,提取其颜色和纹理特征,生成图像的多示例包,然后利用视觉语义模型(visual semantic model,VSM)和TPM获取包特征,最后利用支持向量机(support vector machine,SVM)学习获得查询模型,并对检索结果按相关性大小进行排序。该方法不仅使用关键点映射的思想,充分利用了关键点所属区域的感兴趣信息,而且使用TPM,充分考虑了示例在包中的重要程度,提高了检索准确率。在Corel数据集和SIVAL数据集上的对比实验表明,该方法具有良好的性能。 |
关 键 词: | 计算机应用;区域映射;全概率模型;多示例学习;图像检索 |
Title: | A multi-instance image retrieval method based on region mapping and total probability model |
Author: | WU Jianguo, LI Hui |
Organization: | School of Computer Science and Technology, Anhui University |
Abstract: | Aiming at solving the problem of semantic image retrieval, a multi-instance image retrieval method based on region mapping and the total probability model (TPM) is proposed in this paper. It regards an image as a bag and the regions in the image as instances. Firstly, each image is segmented by the super-pixels method of simple linear iterative clustering (SLIC). Secondly, the key points are found out from the image. Thirdly, the features of color and texture are extracted from those super-pixels which are mapped by the key points, and the multi-instance bag is generated by the feature space. Fourthly, the features of the bag are produced by using the visual semantic model (VSM) and TPM. Finally, the support vector machine (SVM) is used to build the query model and the query results are sorted according to the correlation. The proposed method makes full use of the information of the region which contains the key points using the idea of region mapping, and also fully takes the importance of the instance in the bag into account by using TPM, so that the accuracy of the image retrieval is improved. It is proved that our method has better accuracy compared with other methods through the experiments on the image datasets of the Corel and SIVAL benchmarks. |
Key words: | computer applications; region mapping; total probability model; multi-instance learning; image retrieval |
发表期数: | 2017年8月第16期 |
引用格式: | 吴建国,李辉. 基于区域映射与全概率模型的多示例图像检索方法[J]. 中国科技论文在线精品论文,2017,10(16):1850-1858. |

请您登录
暂无评论