School of Computer Science and Engineering, Vellore Institute of Technology, Chennai Campus, India.
Department of Information Technology, SSN College of Engineering, Chennai, India.
Math Biosci Eng. 2021 Jul 14;18(5):6178-6197. doi: 10.3934/mbe.2021309.
The content-based image retrieval (CBIR) system searches and retrieves the similar images from the huge database using the significant features extracted from the image. Feature integration techniques used in the CBIR system assign static weights to each feature involved in the retrieval process that gives a smaller number of similar images as a result. Moreover, the retrieval time of the CBIR system increases due to the entire database search. To overcome this disadvantage the proposed work introduced a two-level searching process in the CBIR system. The initial level of the proposed framework uses the image selection rule to select more relevant images for the second-level process. The second level of the framework takes the proposed dominant color and radial difference pattern details from the query and selected images. By using color and texture features of the selected images, similarity measure is calculated. The proposed work assigns optimal dynamic weight to the similarity measure of color and texture features using the fruit fly optimization algorithm. This improves the retrieval performance of the CBIR system.
基于内容的图像检索（CBIR）系统使用从图像中提取的显著特征从庞大的数据库中搜索和检索相似的图像。CBIR 系统中使用的特征集成技术为检索过程中涉及的每个特征分配静态权重，这导致结果只显示较少数量的相似图像。此外，由于整个数据库搜索，CBIR 系统的检索时间增加。为了克服这一缺点，提出的工作在 CBIR 系统中引入了两级搜索过程。所提出框架的初始级别使用图像选择规则为第二级过程选择更相关的图像。框架的第二级采用从查询和选定图像中提取的主导颜色和径向差分模式细节。通过使用选定图像的颜色和纹理特征，计算相似性度量。所提出的工作使用果蝇优化算法为颜色和纹理特征的相似性度量分配最佳动态权重。这提高了 CBIR 系统的检索性能。
