Rectangular target extraction for mobile augmented reality applications

We propose a fast and efficient method for localization and rectification of a dominant rectangular region within an image, particularly suitable for mobile Augmented Reality applications. This approach can deal with perspective distortion and high-frequency structures such as text. The resulting image may be used for planar tracking or as input for subsequent image processing tasks. We demonstrate the feasibility of our approach in an extensive evaluation on off-the-shelf mobile hardware. We show its application in two particular use cases, namely planar tracking and mobile visual search. In both cases significant improvements in performance are achieved by using rectified images of the region of interest.