f (u, v) = ^^ i cos mu cos lv + bmt sin mu cos lv m=0 l=0
The weights on each of the terms form the parameter vector p to be optimized. The series is truncated at K so that only a finite number of harmonics are used, in order to limit the search space dimensionality and constrain the space of functions. Different types of surfaces (e.g., open, closed) can be modeled by constraining the parameter values .
We model the image as a Markov random field (MRF) and use a Maximum a posteriori (MAP) probability approach to do region-based classification [27,7,33]. This approach classifies each voxel into a particular tissue type based on its gray level and neighboring voxels in order to ensure a smooth segmentation.
Was this article helpful?