Minimum Description Length segmentation adapted to different image sensors
Philippe Réfrégier, Frédéric Galland, Pascal Martin, Guillaume Delyon, Nicolas Bertaux, François Goudail · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2004
Determining the shape of one or several objects in an image is a fundamental task for many imaging systems. We propose here a general review of new techniques based on the information theory principles, which allows one to determine segmentation techniques without parameter to be tuned by the user. These techniques are quite general since they include, polygonal and spline parametric shape descriptions, level set models of contour and homogeneous partition of images. This approach can take into account the physical nature of the grey level fluctuations and is thus adapted to different new imaging systems. Furthermore, it can lead to fast algorithms (from a few hundred of ms to a few seconds depending on the complexity of the task to perform on 256 x 256 pixel images).