基于视觉语义主题与反馈日志的图像自动标注

为了提高图像标注性能,提出了一种基于视觉语义主题与反馈日志的图像自动标注方法。首 先,提取图像 前景与背景区域,分别进行处理；其次,基于WordNet构建标注词之间的语义关系模型,并 结合概率潜在语义分析(PLSA) 与高斯混合模型(GMM)建立图像底层特征、视觉语义主题与标注

Image automatic annotation based on the visual semantic topics and feedback log

A novel automatic annotation scheme ba sed on the visual semantic topics and feedback logs is proposed in the paper. Firstly,the foreground and background regions of the image are extracted and pr ocessed respectively.Then,the relations among the low-level features,the visua l semantic topics and the key words are built based on the probabilistic latent semantic analysis (PLSA) and the Gaussian mixture model (GMM).After that,the mod el of the semantic relations among the key words is constructed based on the WordNet to improve the annotation performance.Based on the incremental associat ion rule,the relevance feedback technology is further introduced in the propose d annotation model by constructing the feedback log s database.The widely used database of Corel5is used as the test bed,and the ex perimental results show that the new scheme gives better performance than the tr aditional methods.