Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of GraphSearch.
This paper presents an efficient approach to rotation discriminative template matching. A hierarchical search divided in three steps is proposed. First, gradient magnitude is compared to rapidly localise points with high probability of match. This result is refined, in a second step, using orientation gradient histograms. A novel rotation discriminative descriptor is applied to estimate the orientation of the template in the tested image. Finally, template matching is efficiently applied with the estimated orientation and only at points with high gradient magnitude and orientation histogram similarity. Experiments show a higher performance as compared to similar techniques and an efficient implementation of the proposed method in terms of computational complexity.