Skip to main content
. 2021 Apr 3;21(7):2505. doi: 10.3390/s21072505

Figure 2.

Figure 2

Our proposed method pipeline: the video input (a) is partitioned into the simple linear iterative clustering (SLIC) segment (b). The superpixel’s centroid motion is clustered, and feature points match between adjacent frames (d). The adaptive K-means clustering is applied to remove local feature points.