摘要
针对海量图像匹配的速度瓶颈问题,提出一种结合图像SIFT特征和KD树搜索的图像匹配算法,并建立了适应有限内存环境的大型KD树混合存储模式。实验结果表明,该方法能显著提高图像搜索速度和图像库的可扩展性,查准率和查全率也明显高于其他搜索方法。
To deal with the performance bottleneck problem in massive pairwise image matching, an image matching algorithm based on KD-tree is proposed, with a hybrid approach to KD-tree construction under a memory constraint. The experimental results indicate that the proposed method increases image searching speed and the extensibility of image library greatly. It has better performance than other search methods, in terms of precision and recall of image matching.
出处
《计算机时代》
2014年第7期40-42,45,共4页
Computer Era
关键词
图像匹配
特征提取
KD树
近似最近邻搜索
image matching
feature extraction
KD-Tree
approximate nearest neighbor searching
