Stable 2D Feature Tracking for Long Video Sequences

In this paper, we propose a 2D feature tracking method that is stable to long video sequences. To improve the stability of long tracking, we use trajectory information about 2D features. We predict the expected feature states and compute a rough estimate of the feature location on the current image frame using the history of previous feature states up to the current frame. A search window is positioned at the estimated location and similarity measures are computed within the search window. Once the feature position is determined from the similarity measures, the current feature states are appended to the history buffer. The outlier rejection stage is also introduced to reduce false matches. Experimental results from real video sequences showed that the proposed method stably tracks point features for long frame sequences.