论文部分内容阅读
研究利用图像增强、灰度阈值、腐蚀运算等图像操作对新疆草药图像做分割预处理,获取草药图像感兴趣区形状,然后对草药形状进行Radon和解析Fourier-Mellin变换,结合图像形状Hu矩和图像主颜色直方图特征,提取草药图像解析Fourier-Mellin矩形状和颜色混合特征用于图像检索,比较该算法对基于内容的新疆草药图像的检索性能。实验结果表明:解析Fourier-Mellin矩结合草药形状和颜色的混合特征算法具有较好的图像旋转、缩放和平移(RST)不变特性。对基于内容的新疆草药形状图像的检索,其检索性能优于解析Fourier-Mellin矩结合形状特征Hu矩混合特征和结合主颜色直方图特征混合特征的检索效率,其检索结果集前60图的平均查准率达到91.82%。“,”Study the Preprocessing of Xinjiang Herbal medicine image segmentation with the image enhancement ,gray threshold , image corrosion etc. Make the Radon transform and analytical Fourier-Mellin transform on herbal medicine shape, In combination with Hu moments, Extracting the mixed invariant moments of herbal shape for image retrieval , Comparing the image retrieval performance based on the mixed invariant moments, Analytical Fourier-Mellin moments and Hu moments. The experiment results show that the herbal shape mixed invariant moments algorithm has RST(Rotation,Scaling, Translation) characteristics. Its image retrieval performance is superior to the mixed feature of the analytical Fourier-Mellin moments and Hu moments , superior to the mixed feature of the analytical Fourier-Mellin moments and the image’s main color histogram feature on the CBIR(Content-Based Image Retrieval). The average precision rate reaches 91.82%at the top of 60 images of the query results.