Robust Computational Vision

Abstract

This paper presents a paradigm for formulating reliable machine vision algorithms using methods from robust statistics. Machine vision is the process of estimating features from images by fitting a model to visual data. Computer graphics programs can produce realistic renderings of artificial scenes, so our understanding of image formation must be quite good. We have good models for visual phenomena, but can do a better job of applying the models to images. Vision computations must be robust to the kinds of errors that occur in visual signals. This paper argues that vision algorithms should be formulated using robust regression methods. The nature of errors in visual signals will be discussed, and a prescription for formulating robust algorithms will be described. To illustrate the concepts, robust methods have been applied to three problems: surface reconstruction, image flow estimation, and dynamic stereo.