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.
