From a Robust Hierarchy to a Hierarchy of Robustness

Abstract

An attempt is made to present a (somewhat personal) history of how at the Com-puter Vision Laboratory in the late 1980's adopting robust techniques from statis-tics naturally arose from a quest for better multiresolution image analysis algo-rithms. While today these robust techniques are routinely used in the vision com-munity, their rapid dissemination was in no small measure due to the unfettered research atmosphere which characterized the lab. Beside trying to record an in-stance of interdisciplinary research, a few technical issues (most of which were yet to be understood then) are also discussed.