Non-rigid Joint Segmentation and Registration Using Variational Approach for Multi-modal Images
- Resource Type
- Authors
- Lavdie Rada; Mazlinda Ibrahim; Adela Ademaj; Ke Chen
- Source
- Advances in Intelligent Systems and Computing ISBN: 9783030665005
- Subject
- Active contour model
Computer science
business.industry
ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION
Image registration
Curvature
Regularization (mathematics)
Modal
Computer Science::Computer Vision and Pattern Recognition
Medical imaging
Computer vision
Segmentation
Artificial intelligence
business
Joint (audio engineering)
- Language
Segmentation and registration are two common problems in medical imaging in particular and computer vision in general. These two problems have been studied substantially in the past two decades and often required as simultaneous tasks. Combination of these tasks in a single framework has proven to yield better results in terms of accuracy. In this paper, a model for the joint segmentation and registration based on the variational approach is presented for non-rigid multi-modal. The model is based on an active contour without edges for the segmentation task, normalized gradient fields, and the linear curvature model for image registration. The discrete functional of the new joint model is optimized using coarse to fine registration. The model is evaluated on synthetic and medical images and compared with the existing model. The proposed model is comparable based on the two evaluation criteria used.