Nonrigid medical image registration technique as a composition of local warpings
2004
We introduce a new technique for nonrigid image registration based on the composition of local deformations. The warping model is analyzed in order to guarantee continuity, differentiability and a one-to-one transformation by constraining the parameters of the nonlinear spatial transformation. A genetic algorithm solves the model by global optimization, handling constraints, and maximizing the normalized mutual information. The composition of local transformations goes throughout several levels of resolution, from coarse to fine. The performance of our technique was tested in synthetic and real medical images. The proposed method was always able to improve the similarity criterion between image pairs, demonstrating the robustness of the method for several modalities of images.
Keywords:
- Machine learning
- Normalization (statistics)
- Image warping
- Global optimization
- Robustness (computer science)
- Artificial intelligence
- Genetic algorithm
- Nonlinear system
- Computer vision
- Mathematical optimization
- Mutual information
- Image registration
- Mathematics
- Differentiable function
- normalized mutual information
- Pattern recognition
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
46
References
11
Citations
NaN
KQI