期刊论文

中文

图书情报工作

0252-3116

2013

57

5

120-123

教育部人文社会科学研究规划基金项目(11YJAZH040)

本校为第一机构

针对图像的特性,给出一种图像的二元视觉语言模型,在此基础上提出一种新的图像语义挖掘方法。该方法将每幅图像表示成一个由视觉单词构成的矩阵,通过计算每个视觉单词的权重,按照权重的大小对视觉单词进行选择,利用选择后的视觉单词集合,构建图像的视觉语言模型;之后,按照贝叶斯公式,建立基于视觉语言模型的图像语义挖掘方法。实验结果表明,该方法在图像语义描述能力和区分性方面是有效的,能充分反映人对图像内容的理解,具有很好的应用价值。

For the features of images, this paper proposes a binary visual language model of the image and a new image semantic mining method. Firstly, it presents each image as a matrix constituted by the visual words, selects the visual words by calculating their weighting value and develops a new image visual language model. According to the Bayesian formula, this paper establishes the image semantic mining method based on visual language model. The experimental result shows that this method is effective in the descriptive ability and distinction of image semantic, fully reflects the understanding of ...