The viper algorithm: A new approach for detection using 2D-adaptive sampling

This paper describes a new technique of irregular and adaptive sampling developed to process still images containing objects of interest over a non interesting background. The method allows the identification of interesting elements and is able to choose the sampling interval more adequate to obtain a correct description of them. The not interesting background is roughly sampled and this particular aspect of the method is an important advantage. Objects are first detected by a preliminary sampling phase and then a resampling process begins, until a convergence test in the Fourier domain is satisfied and stops the process. Examples of applications are given.