Object detection using a novel 2-D adaptive sampling strategy

This paper describes a novel algorithm especially devoted to detect "interesting" objects defined as a set of close measured values that spring out from a uniform or textured background. In particular such an algorithm has been developed with the purpose to define a movement strategy for a remotely piloted rover useful in applied geophysics. The method could be represented by a Viper that moves with a random trial over the image with the purpose to detect the presence of an object emerging from the background, and to adapt the sampling interval to the shape of the detected object. Simulations on real geophysics images show good results with reduced algorithm complexity.