Hybrid Indexes to Expedite Spatial-Visual Search

TL;DR

This paper introduces hybrid index structures that combine spatial and visual features to improve the efficiency and accuracy of geo-tagged image searches within specific geographic areas.

Contribution

It proposes novel hybrid index structures that evaluate spatial and visual features together, addressing inaccuracies in both to enhance search performance and relevance.

Findings

Hybrid indexes outperform baseline indexes in search speed.

Hybrid indexes improve result relevance accuracy.

Experiments on real datasets validate the effectiveness of the proposed structures.

Abstract

Due to the growth of geo-tagged images, recent web and mobile applications provide search capabilities for images that are similar to a given query image and simultaneously within a given geographical area. In this paper, we focus on designing index structures to expedite these spatial-visual searches. We start by baseline indexes that are straightforward extensions of the current popular spatial (R*-tree) and visual (LSH) index structures. Subsequently, we propose hybrid index structures that evaluate both spatial and visual features in tandem. The unique challenge of this type of query is that there are inaccuracies in both spatial and visual features. Therefore, different traversals of the index structures may produce different images as output, some of which more relevant to the query than the others. We compare our hybrid structures with a set of baseline indexes in both performance and result accuracy using three real world datasets from Flickr, Google Street View, and GeoUGV.