A numerical comparison between Feature Correlation and Phase Correlation for motion estimation relative to sea bottom

The article compares the performance of Feature Correlation and Phase Correlation algorithms for the motion estimation of an underwater vehicle. While feature correlation relies on the extraction and matching of features templates, Phase Correlation is a featureless approach, relying instead on the overall properties of the image.Three versions of Feature Correlation algorithm and two different implementations of the Phase Correlation algorithm are compared, on data collected by the HCMR Thetis submersible during the Nautilos Project in 2004; since the mission focused on different research priorities, the recordings are suboptimal regarding sea bottom mosaicking, featuring motion blurs, roll sweeps, geometric distortions and optical axis non-perpendicular to the bottom.The results show the effectiveness of feature correlation in dealing with blurred images, while in terms of computational time, both implementations of the phase correlation algorithm outperform any of the feature correlation algorithms by a large margin.