图像重排序中与查询相关的图像相似性度量

现今的图像搜索引擎主要利用图像周围文本信息为图像排序,根据图像内容重排序可以进一步提高搜索性能。图像相似性的度量对重排序算法的性能至关重要。然而已有的相似性度量没有考虑针对不同的查询,图像的相似性应该不同。提出一种与查询相关的相似性度量方法,将基于全局特征的相似性,基于局部特征的相似性,以及视觉单词同时出现率融合到一个迭代算法中,挖掘出与查询相关的图像信息,计算图像相似性。在Bing图像搜索引擎上的实验结果证明本文提出的相似性度量方法优于基于全局特征,局部特征,或它们线性组合的相似性。

Query Dependent Visual Similarity in Image Search Reranking

Recently image search engines mainly base on associated textual information. Image reranking is an effective approach to refine the initial text-based search result by mining the visual information of the returned images. And the estimation of visual similarity is the fundamental factor in reranking methods. However, the existing similarity measures are independent of the query. This paper proposes a query dependent method by incorporating the global visual similarity, local visual similarity and visual word co-occurrence into an iterative propagation framework. Then it embed the query dependent similarity into random walk rereanking method. The experiments on a collected Live Image dataset demonstrate that the proposed query dependent similarity outperforms the global, local similarity and their linear combination.