Shape Analysis

Introduction This paper presents an algorithm for generating statistical shape models by addressing it as a correspondence problem of finding the parameterization of each shape in the training set, instead of manually annotating a set of “landmark” points for each image in the training set. The authors demonstrate the robustness of the algorithm by applying it to a variety of training image sets - infarcts, kidneys, knee cartilages, hand outlines, hip prostheses, and left ventricles. The proposed minimum description length model leads to good compactness, specificity and generalizes well, outperforming the contemporary gold standard - manual landmarking. Moreover, the authors also show that this model can be extended to work with 3-D images. ...

Introduction This paper presents an algorithm for modeling rigid objects in the presence of noise, clutter, and occlusion and overcomes the problems facing the contemporary algorithms - sacrificing specificity to accommodate variability. The proposed models permits deformations only consistent with the class of objects it represents. The authors demonstrate the robustness of the algorithm by applying it to a variety of training image sets - resistors, heart, hand, and worm models. ...