Robust feature tracking in underwater video sequences

The paper proposes a robust feature tracker based on an efficient outlier rejection scheme, suitable for feature tracking in subsea video sequences. We extend the Shi-Tomasi-Kanade scheme (C. Tomasi and T. Kanade, 1991; J. Shi and C. Tomasi, 1994) by introducing a technique for rejecting spurious features. We employ a simple and efficient outlier rejection rule, called X84, and prove that its theoretical assumptions are satisfied in the feature tracking scenario. Experiments with synthetic and real subsea sequence confirm that our algorithm locates and discards unreliable features accurately and consistently, and tracks good features reliably over many frames. We also illustrate quantitatively the benefits introduced by the algorithm with the example of fundamental matrix estimation