1: \begin{abstract}

2: Sampling from distributions of implicitly defined shapes enables

3: analysis of various energy functionals used for image

4: segmentation. Recent work \cite{Chang2011} describes a

5: computationally efficient Metropolis-Hastings method for accomplishing this task. Here, we

6: extend that framework so that samples are accepted at every iteration

7: of the sampler, achieving an order of magnitude speed up in

8: convergence. Additionally, we show how to incorporate topological

9: constraints.

10: \end{abstract}

11: