基于基分类器的图像语义标签自动排序方法ISLR-BV

Image Semantic Label Automatic Ranking Method ISLR-BV Based on Base Classifier

针对当前图像语义标签的无序性问题,提出了一种基于基分类器加权投票的图像语义标签自动排序方法 ISLR-BV。该方法综合考虑图像的显著性区域内容以及图像的语义上下文信息,借助RAPC-W进行多标签数据集的转化,并在RAPC-W转化后的数据集上进行基分类器的学习,借助得到的基分类器对图像的语义标签进行加权投票,以此来决定每个语义标签与图像的相关程度,实现图像语义标签的有序排列。在数据库NUS-WIDE上的实验结果表明,在一定程度上提高了图像语义标签排序的准确度。

In view of the current image semantic tagging disorder problem, this paper proposes a image semantic label ranking method LSLR-BV based on the weighted voting of base classifier. The method considers the signifi- cant regional content of image and the image semantic context information, conducts the base classifier learning on the base of the RAPC-W conversion data set, takes the weighted voting on the image semantic labels with the help of the base classifier, in order to determine the related degree between the each semantic labels and the im- age, implements the ordered arrangement of the image tag order. The experimental results on the database of NUS-WIDE show that the. method is effective and feasible.