Initialization Free Segmentation of Complex Shaped Objects
Nalan Karunanayake, Stanislav S. Makhanov · 2019
This paper proposes a new, initialization free segmentation method based on dynamic active agents interacting with the from the image edges. The agents are able to segment complex shaped single and multiple objects characterized by deep and narrow concavities (1 pixel wide) a low contrast and a high level of noise. Numerical tests against conventional methods show an advantage of the proposed algorithm.