Image Segmentation by Image Foresting Transform with Boundary Polarity and Shape Constraints

Image segmentation, such as to extract an object from a background, is very useful for medical and biological image analysis. In this work, we propose new segmentation methods for interactive segmentation of multidimensional images, based on the Image Foresting Transform (IFT), by exploiting for the first time non-smooth connectivity functions (NSCF) with a strong theoretical background. The new algorithms provide global optimum solutions according to an energy function of graph cut, subject to high-level boundary constraints (polarity and shape). Our experimental results indicate substantial improvements in accuracy in relation to other state-of-the-art methods, using medical images by allowing the customization of the segmentation to a given target object.

Object segmentation by Oriented Image Foresting Transform with connectivity constraints

Global properties, such as connectivity, shape constraints and boundary polarity, are useful high-level priors for image segmentation, allowing its customization for a given target object. In this work, we introduce a new method called Connected Oriented Image Foresting Transform (COIFT), which provides global optimum solutions according to a graph-cut measure, subject to the connectivity constraint in Oriented Image Foresting Transform (OIFT), ensuring the generation of connected objects, as well as allowing the simultaneous control of the boundary polarity. While the use of connectivity constraints in other frameworks, such as in the min-cut/max-flow algorithm, leads to an NP-Hard problem, COIFT conserves the low complexity of the OIFT algorithm. Experiments show that COIFT can considerably improve the segmentation of objects with thin and elongated parts, for the same number of seeds in segmentation based on markers.