Abstract:

In the research area of Content Based Image Retrieval(CBIR), aiming at extracting the interesting image content to improve retrieval results, this paper proposes an algorithm based on dynamical region pair matching through selecting images by users, uses the Fisher prediction to obtain the dynamical threshold to explore the “interesting” components of the features in the region pair match, adopts those interesting components of the features and the region pairs to retrieve the similar images. The experimental results show the method is effective.

Key words: dynamical threshold selection, “interesting” components of features, content-based image retrieval

摘要： 在基于内容图像检索研究中，针对如何提取用户感兴趣的内容以提高检索效果的问题，该文根据用户选择的相似图像自动获取“感兴趣”区域，使用Fisher判别进行区域中“感兴趣”特征分量阈值的动态选择，并以该“感兴趣”特征分量及“感兴趣”区域作为用户感兴趣的内容进行图像检索。试验结果表明了该方法的有效性。

关键词: 动态阈值选择, “感兴趣”特征分量, 基于内容图像检索

CLC Number: 

• TP18