基于可视特征度量距离的图像检索

Image indexing based on distance metrics of visual features

研究预滤波、二进制集的划界、降维和查询优化的距离计算 ,并在 40 0 0幅测试图像集上 ,采用实验方法对高维直方图的索引策略与优化查询的距离计算进行了评估 .研究表明 ,非二次形式的距离度量与二次形式的距离度量一样有良好的性能 .此外 ,二进制集是直方图的一种有效而简捷的替代方法 ,其检索性能足以满足第一遍查询 .

Pre-filtering, binary set bounding,dimensionality reduction and query optimized distance computation are investigated,and the retrieval strategy for high dimensional histograms and the query optimized distance computation by experiments on the test set of 4?000 images are evaluated. The research showed that the non-quadratic form distance metrics perform as well as the quadratic form metrics. Furthermore,binary sets are an efficient and compact alternative to histograms, and their performance suffices as a first pass query.