Machine vision is a branch of artificial intelligence aimed at extracting useful information from images. Vision techniques have typically been applied to images acquired by cameras, but recent research efforts have seen novel algorithms work with paintings. Promising results have been obtained which cast a new light on the debate regarding the way traditional paintings were created. The author believes that the process of seeking answers to this controversial debate cannot be driven by predetermined stances and superficial investigations. In many cases, only a careful scientific analysis supported by the guidance of expert art historians may lead to safe, conclusive answers. Computer vision and projective geometry may provide some of the basic tools to effectively tackle the problem. This document illustrates examples of cases where superficial analyses have lead to incorrect conclusions as well as cases where rigorous vision techniques have provided interesting insight on how a painting (or a portion of it) was generated.
In Tech. Digest of the Annual Meeting of the Optical Society of America (OSA).