Photograph Indexing and Retrieval using Combined Geo-information and Visual Features

This paper proposes a new method to automatically index searches for relevant images using geo-coded information. Photographic images are labeled with their GPS (Global Positioning System) coordinates and date/time at the moment of capture and this date is then utilized to create two layer spatial and temporal indexes for image searches. A simulation performed to estimate the effectiveness of the proposed method found that the combination of those indexes and visual features improving average recall by up to 59% and average precision by up to 49% compared to the use of visual content alone.