Estimation of Geographic Information from a Single Image

Abstract

With the internet rapidly turning into a hub of available information resource, social networking sites have seen an unprecedented increase in their popularity during the past decade leading to millions of users joining from around the world. Some of these social networking sites are dedicated to only image and video content based experience sharing. Flickr.com is one of these popular image sharing sites owned by yahoo group where each day millions of images are shared online from people around the world. Some of these images are even available with a very precise information about their location. Considering this, it is becoming more and more feasible for us to start utilizing this information in form of images for enhanced future of the computer vision field. These available images with known locations can be used to determine or predict the location of images whose location is not already known. This report covers the details about a proposed system which can be used to estimate the geographic information of images with unknown origin based on the available data of images with precisely known locations. Different high and low level image processing techniques will be utilized in this regard which are discussed in detail in the coming chapters of this report. The idea is to extract useful features from images in training data with known location; and compare them with features of test data images whose location is to be estimated. By doing so, we will be able to find an approximate match of test data image in training data images, based on this match; some estimations can be made about the location of test data images.