A Bayesian approach to PET reconstruction using image-modeling Gibbs priors: implementation and comparison

A hybrid framework for 3D medical image segmentation

“…However, in (Geman and Geman, 1984;Derin and Cole, 1986;Lakshmanan and Derin, 1989) the segmentation performance is limited by the fact that the pairwise Gibbs prior model contains only incomplete boundary information. In (Chan et al, 1997(Chan et al, , 1998, Chan used a high order Gibbs prior model which integrates the region and the boundary information to reconstruct medical images from 2D projections. The advantage of this approach is to use a Gibbs distribution that integrates the boundary information instead of previous Gibbs prior models that only considers the pairwise pixel similarity.…”

Hybrid segmentation framework for 3D medical image analysis

“…Although Gibbs Prior models and deformable models have been used separately in segmentation or other image processing tasks before [6,7,8,9,21,42,43], this is the first effort to combine them into one hybrid framework for 3D medical image segmentation. Compared to previous work, our method has the following advantages:…”

Image Segmentation Based on the Integration of Markov Random Fields and Deformable Models

“…Recently, there have been several hybrid methods such as the ones developed in [5,6,15,16,17]. In [18], Zhu et al develop a unifying framework that generalizes the deformable models, region growing, and prior matching approaches.…”

“…[8,9,19] used MRF or Gibbsian models that include nearest neighbor correlation information, but do not use object boundary information. [6] and [7] defined higher order neighborhoods in MRF image models that model both region and boundary object information based on theoretical results from [20,21].…”

“…Chan et al apply the Gibbs prior model in [6] and [7] for image processing based on the work of Besag [8] and Geman and Geman [9]. Compared to the affinity operator, the Gibbs prior models can incorporate much more information.…”