INVARIANT SNAKES AND INITIALIZATION OF DEFORMABLE MODELS
Gilson Antonio Giraldi, Antonio A. F. Oliveira · International Journal of Image and Graphics · 2004
Parametric snake models are deformable models very useful for boundary extraction and tracking. Despite their abilities, the non-invariance of the internal energy under affine transformations and the non-convexity of the model energy are known limitations for most of these methods. The former has been addressed in the context of active shape models. The latter can be addressed through Dual contour approaches as well as an automatic procedure to initialize the model closer to the desired boundary. In this paper we first analyze the effects of reparameterization for invariant snake models through the Lie Group Theory. In the second part of this work, we present our Dual-T-Snakes model and a methodology based on the T-Snakes framework to initialize deformable models. In the experimental results we explore and combine these techniques for segmentation of medical images.