Abstract:

The complex medical images are not always described the parametric method with prior probability, which leads to the difference between the actual physical model and the basic hypothesis of the model, namely, the “model mismatc”. The method of CT image segmentation based on the multimodal function optimization was proposed. A density model of the nonparametric orthogonal polynomials for image data was proposed, and the novel particle swarm optimization method was used to solve the multimodal function optimization problem. On the basis of the heuristic optimization search, the novel method was successfully applied in multimodel function optimization. Some experiments about artificial data and real CT images were tested. These results show the validity of the algorithm in image segmentation.

Key words: nonparametric density estimation, medical image segmentation, orthogonal polynomial, tensor product, particle swarm optimization