An efficient autofocus algorithm for a visible microscope on a Mars lander

The limited depth of focus of microscope optics precludes the observation of a three-dimensional surface in a single view. For an efficient operation of a planetary microscope observing a highly irregular surface, an autofocus algorithm must detect partially focused views and merge them into a completely focused image (extended depth of focus). In this paper we investigate the suitability of different algorithms for autonomous image focusing onboard a Mars lander. We show that all methods under study, except for the Variance method, produce reasonable results and that their performance depends (amongst other factors) on sample morphology. For reasons of simplicity and reduced computational requirements we favor Gradient and Roberts methods and present an efficient implementation thereof.