Advanced System for Identifying Ancient Photographs Using Contemporary Tagged Images

Abstract

Huge collections of imagery on the Internet have inspired a wave of approaches to location recognition, the problem of determining where a photo was taken by comparing it to a database of images of previously seen locations from the past few years. Due to this, as excitement in these areas increases, a world-scale location recognition engine from all the geo-tagged pictures from online photo collections, such as Flicker, Instagram, and street view databases from Microsoft and Google. Matching modern historical images to old ones requires a special effort when dealing with historical photos. The performance of many algorithms is good on modern pictures but is not very efficient on old historical photos. A novel approach to place identification, taking pictures taken in different weather conditions, illuminations, and positions, is proposed in this paper, expecting more accuracy than other existing systems.