Hi, I am working on research about 3D medical segmentation with Chan-Vese. I suggest to use swarm optimization to find the optimal parameters of the segmentation method, but the result is not good enough(MSE, PSNR). can any one suggest any thing to enhance the result (MSE, PSNR)?
Quality of segmentation is about comparing with ground truth in terms of https://en.wikipedia.org/wiki/Confusion_matrix . On the right side of wiki-page you will see some formulas for quality of segmentation. PSNR and MSE never used in segmentation's evaluation.
You can try to ehnance image contrast before the active contour segmentation. In this way your segmentation should be more accurate
Chan-Vese is non-convex, therefore you will normally find suboptimal solutions. have you considered to use the convex relaxated version of Mumford-Shah and globally optimize for example with continuous-cut?
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Computer Graphics and Animation
- 3D