An Investigation of Using K-d Tree to Improve Image Retrieval Efficiency
Yunshuang He, Guojun Lu, Shyh Wei Teng · 2002
Content-based image retrieval problem is in essence to determine distance or similarity between multi-dimensional image feature vectors. Linear comparison will be too slow when vector dimensions and image database are large. We investigate the use of K-d tree in image retrieval based on color histograms and found that K-d trees improve retrieval efficiency significantly. We evaluate the impact of leaf node size and the number of images required to retrieval on the retrieval performance. Finally, we discuss the relationship between vector dimensions and retrieval efficiency. 1.