AI Summary & Key Takeaways
AI InsightThrough experiments, it is shown that the proposed model finds the aimed object boundaries successfully for smooth or challenging oscillatory topological structures.
📌 3 Key Takeaways
Abstract
Accurate boundary determination and segmentation of an object of interest in an image is a difficult image segmentation task. In this paper, we propose a new model which improves the old selective segmentation of Rada et al. [23] by combining two penalization and two fitting terms. To better deal with oscillatory boundaries, a H^1 weighted length term and L^2 Lebesgue measure have been employed as penalization terms, whereas the fitting terms consist of a region-based and area fitting term. The model has the same speed as the previous one-level set interactive segmentation model by Rada et al. [23] and much faster compared to previously dual-level set models by Rada et al. [25, 24] by having the same segmentation accuracy and reliability. On the other hand, the model shows a better performance while dealing with irregular and oscillatory object boundaries compared to Rada et al. [25] model. Further comparison with segmentation algorithms of the same nature, such as the Nguyen et al. [21] method, shows that the proposed model shows the same or improved performance for object segmentation with transparent boundaries or inhomogeneous intensity of the aimed object. Through experiments, it is shown that the proposed model finds the aimed object boundaries successfully for smooth or challenging oscillatory topological structures.