Abstract

We present a novel variational approach to top-down image segmentation, which accounts for significant pro jective transformations between a single prior image and the image to be segmented. The pro- posed segmentation process is coupled with reliable estimation of the transformation parameters, without using point correspondences. The prior shape is represented by a generalized cone that is based on the con- tour of the reference ob ject. Its unlevel sections correspond to possible instances of the visible contour under perspective distortion and scaling. We extend the Chan-Vese energy functional by adding a shape term. This term measures the distance between the currently estimated sec- tion of the generalized cone and the region bounded by the zero-crossing of the evolving level set function. Promising segmentation results are ob- tained for images of rotated, translated, corrupted and partly occluded ob jects. The recovered transformation parameters are compatible with the ground truth.