摘要
针对基于内容的图像检索系统的检索效率和精度的不足,提出了综合语义和轮廓特征的图像检索方法。以拐点作为控制点对图像的轮廓进行精确分段,利用边界跟踪法对图像进行轮廓特征提取,并以图像的语义和底层的轮廓特征作为图像检索的综合指标,将图像的主观语义和底层特征融合起来,提高了图像底层特征和高层语义之间的联系。通过对不同类型的图像进行检索,实验结果证明该算法对复杂图像检索的效率高、精度高,并具有稳定的检索性能。因此,具有很好的发展趋势。
To deal with the defects of the efficiency and accuracy of CBIR, this paper focuses on a new method for image retrieval using both semantic information and figure features. A Inflexion - Based Method for Precise Contour Segmentation is proposed. Edge tracing is used for the feature extraction of the figure in this paper and a new method based on figure and semantic information for image retrieval system is built to fill up the difference between figure and the semantics. Experiments with different types of images show that the approach can improve the efficiency and precision of image retrieval and has steady image retrieval performance.
出处
《计算机仿真》
CSCD
2008年第5期195-198,共4页
Computer Simulation
关键词
图像检索
语义
轮廓特征
拐点
Image retrieval
Semantic information
Figure features
Inflexion
