image segmentation

Hi, I'm looking for some who can implement a method of segmentation that works reasonably good for noisy volumetric images (CT scans of foot. 512 px x 512 px x 512 px). Input is contiguos array of bytes representing voxel intensities arranged in 3D space. Output should be array of the same length containing label for each voxel. It is necessary to label disjoint regions containing the same tissue with the same label. Method does not have to be fully automatic, but we want to limit it to one-two human operator actions (Modification of region growing algorithm variant should be fine here, but vanilla region growing does not work well). In case of doubt we want the algorithm to tend to oversegment. If you have algorithm you know will work well don't hesitate to use it straight away, what is necessary for me is references to articles/books in documentation. Yes, clear documentation is a must for me. To integrate with my library you will have to have account on Bitbucket, Nvidia SDK installed on your machine, CMake 3.8 and Visual Studio 2015. It is much more important for the method to be fast than 100% correct. If output has to be fixed by human operator with lets say 1-2 actions it should be all right. Specific technologies required: C/C++, NVidia, Cmake