Improving Quality of Digital Images of Art in Museum Collections

Abstract

In this paper, we present a study in which we analyze digital image quality corrections performed by an expert operator at the National Gallery of Art (N.G.A.) in Washington, D.C., and we propose a framework to semi-automatically improve the quality of digital images in museum collections. The work presented has two goals: (1) to explore ways to facilitate the color image correction process, and (2) to gain a better understanding of it. We analyze the expert's correction process (i.e., operations and workflow), and compare changes in contrast and luminance for original and corrected images selected from two different collections (Impressionist and Dutch/Flemish paintings). Results of the study suggest that, although corrections depend on each individual image, it is possible to find patterns in the way that similar images are corrected. Therefore, the proposed framework is based on the assumption that images can be placed in categories (images within a category are more visually similar than images across categories), and that correction patterns can be learned and applied semi-automatically (i.e., under the supervision of an expert operator) for different categories.