基于SIFT特征的文物图像检索

Cultural relic image retrieval based on SIFT features

目的研究解决西北大学数字博物馆中文物图像的检索问题。方法提出了一种基于尺度不变特征变换(SIFT)特征的文物图像检索方法。结果首先提取图像的SIFT特征,并利用主成分分析方法进行特征降维;在对SIFT特征进行匹配时,利用近似最近邻搜索算法检索匹配特征点;最后,使用局部几何约束控制误匹配特征点,以进一步提高检索准确度。在文物图像库上进行实验,表明该方法是有效的。结论基于SIFT特征的图像检索提高了文物图像检索系统的性能。

Aim To study the cultural relic image retrieval problems in digital archaeology museum of Northwest University.Methods An image retrieval method based on scale-invariant feature transform（SIFT） features is proposed.Results Firstly,SIFT features of images are extracted and Principal Component Analysis method is adopted to reduce the dimensionality;then best bin first search method is employed in SIFT feature points matching;finally,for further improving the retrieval accuracy,local geometric constraint is used to control mismatching feature points.The experimental result on cultural relic image database shows that this method is feasible.Conclusion The proposed image retrieval method,which is based on SIFT features,improves the performance of cultural relic image retrieval system.