Implement of following paper:
"L. Ben Boudaoud, A. Sider and A. Tari, "A new thinning algorithm for binary images" 2015 3rd International Conference on Control, Engineering & Information
Technology (CEIT), 2015, pp. 1-6, doi: 10.1109/CEIT.2015.7233099.
Compare with ZS algorithom：
T. Y. Zhang and C. Y. Suen. 1984. A fast parallel algorithm for thinning digital patterns. Commun. ACM 27, 3 (March 1984), 236–239.
ZS Algorithm
Algorithm description:
For the first iteration , if the edge of point P1 satisfies the following conditions, it will be deleted P1：
(1) 2 ≤ 𝐵(𝑃1) ≤ 6
(2) 𝐴(𝑃1) = 1
(3) 𝑃2×𝑃4×𝑃6= 0
(4) 𝑃4×𝑃6×𝑃8= 0
𝐴(𝑃1) When traversing the 8 neighbors of 𝑃1，the number of values changing from 0 to 1 (white to black)；
For the second iteration, if point P1 satisfies the following conditions, will deleted P1：
(1) 2 ≤ 𝐵𝐵(𝑃𝑃1) ≤ 6
(2) 𝐴𝐴(𝑃𝑃1) = 1
(3) 𝑃𝑃2×𝑃𝑃4×𝑃𝑃8= 0
(4) 𝑃𝑃2×𝑃𝑃6×𝑃𝑃8= 0
The New Thinning Algorithm
Algorithm description：
(1) Use the same 8-connected neighborhood as the ZS algorithm.
(2) 𝑃1 is the pixel to be deleted.
(3) 𝐵(𝑃1) = ∑ 𝑃𝑖𝑖9𝑖𝑖=2 is the number of 1 (black grid) in 𝑃1、𝑃2、𝑃3 … 𝑃8.
(4) 𝐶(𝑃𝑃1) =¬𝑃2∧ (𝑃3∨𝑃4) + ¬𝑃4∧ (𝑃5∨𝑃6) + ¬𝑃6∧ (𝑃7∨𝑃8) + ¬𝑃8∧ (𝑃9∨𝑃2) is the "1" in the 8-connected neighborhood of the number of connected 8-components.
Result
Image
ZS algorithm
---
---
---
------
TR
ET(s)
SM
CM
Duck
0.9985342
34.235636
4
4
Image
new thinning algorithm
---
---
---
------
TR
ET(s)
SM
CM
Duck
0.998889
18.152677
10
6
TR：Thinning rate, The degree of thinness of the image.
ET：Execution Time ,The real time cost ,the execution time of thinning process in second.
SM：Sensitivity Measurement, Measure the noise immunity performance of the contour bobbin.
CM：Connectivity Measurement, Measuring the connectivity of the output skeleton computes disconnected lines or discrete points that should not appear,
which makes what would otherwise be the same object be considered different objects.