A method and apparatus to calibrate a printer which do not depend on geometric assumptions on the printed dots. The method takes account of the fundamentally probabilistic nature of the process of printing any dot on some types of printers such as laser printers. It allows estimating the probability distribution of what gets printed at each pixel depending on the local configuration of dots to be printed in such a way that only a limited number of tests need to be done, while no artificial cutoff is introduced on the size of local configurations to be taken into account. The calibration method can be applied to many problems encountered in digital printing, and permits computing printer dependent calibration of a dithering method without need for further direct measurements.