Automatic extraction of circular objects in a digital image

This presentation has two purposes. First to present the higher-order active contour (HOAC) framework for modelling and identifying networks or sets of objects of specific shapes in a digital image. Second, to apply this framework to the extraction of a network of circular objects as for example tree crowns in zenithal images of regular tree plantations. Other applications may feature detection/extraction of biological cells on microscopical images, craters, missile silos, etc.

To achieve this specific goals we set the parameters to modeling a ‘gas of circles’, the set of regions in the image domain consisting of an unknown number of circles with approximately fixed radius and short range repulsive interactions ; second, we apply the presented model on very high resolution remotely sensed images to extract tree crowns.The approach differs from most other methods for incorporating prior geometric knowledge into active contours and for not being based upon perturbations of a reference set of regions. In consequence, it can detect multiple instances of an entity at no extra cost, a critical requirement for the current application.

Speaker

P. Horvath, University of Szeged, Hungary

Organisers

• French Institute of Pondicherry (Dr. Pierre Couteron)
• Pondicherry University (Prof. Kuppuswamy)

Venue

Pondicherry University. Seminar Hall of the dept. of Computer Science, Ramanujan School of Mathematics and Computer Science, Science & Humanities Block II

Time

15h00