A new algorithm with adaptive feature extraction and feature selection simultaneously is proposed to improve the performance of Content-based Image Retrieval( CBIR) . Semantic gap between low-level visual features and high-level semantic information is reduced by synchronization in feature extraction and feature selection. A parameterized wavelet is used to improve accuracy of image details. Mother wavelet function of color histogram feature is optimized and interval parameters are quantified using multiple gravity search algorithm. Experimenal results on 1 000 images searched by Corel show that compared with the most relevant algorithm, fusion algorithm of Gravitational Search Algorithm and Support Vector Machine(GSA-SVM),fusion algorithm of Fuzzy Color Histogram and Fuzzy String Matching(FCH-FSM),the retrieval accuracy is higher,and the and average time consumption is less.%为提高基于内容的图像检索( CBIR)算法的检索性能,提出一种同时进行自适应特征提取和选择的CBIR算法。该算法通过同步特征提取和选择,减少低级视觉特征和高级语义之间的语义差距,使用参数化小波提高图像细节的准确度,利用混合引力搜索算法优化颜色直方图特征中母小波函数和量化间隔参数。在Corel收集的1000幅图像上的实验结果表明,相比最相关特征算法、引力搜索算法和支持向量机的融合算法、模糊颜色直方图和模糊字符串匹配的融合算法,该算法的检索精度较高,平均耗时较少。
展开▼