Active contours for selective object segmentation

Selective object segmentation is a difficult, but often required image analysis task which has numerous applications in various fields such as in biology, satellite imaging or video surveillance. We present a family of active contours designed to segment objects with predefined size and simple shape characteristics using the ratio of area and perimeter, and the second central moment of the contour. These priors allow targeted selection of objects by their size and shape with certain flexibility. A novel data model is introduced based on local regions of the contour points. The proposed methodology is tested on synthetic and fluorescent microscopic images containing a mixture of genetically modified bacterial populations. The method was capable of selecting objects by their size and shape.