KRLT Comparison

With two direction generating algorithm, it is natural to ask which one could achieve a better result. After conducting tests on some nerve photo, we came to the conclusion that the second mode outperforms the other. One of the test cases is elaborated as follows.

From above two figures, it is obvious that the second algorithm could produce a more desirable result. Most of the nerve could be tracked clearly in the second one, while the first image contains some weird paths that are not consistent with the original image.

We hypothesized that the path in the first image is easy to step out of the nerve because of inaccurate direction, while directions generated by the second one algorithm are valid enough. To verify this, we zoom in at the central region where the two images differ from each other. The direction result comparison is shown below.

The blue line segment indicates the direction for its corresponding red start pixel. As illustrated in the figure, the directions at this patch produced by first algorithm are not quite consistent with the real picture. While the nerve actually goes from left upper corner to right down corner, the plot fails to show this trend. As a contrast, the second figure give us a valid result.