Lay your sleeping head, my love,
Human on my faithless arm;
Time and fevers burn away
Individual beauty from
Thoughtful children, and the grave
Proves the child ephemeral:
But in my arms till break of day
Let the living creature lie,
Mortal, guilty, but to me
The entirely beautiful....
—W. H. Auden
Garth Herrick is a realist painter working in the Philadelphia area, and his Human on my faithless arm, in a private collection was executed in two parts. The background was based on a photograph of the architectural scrollwork and the self-portrait under separate studio conditions, including lighting. The following is based on:
Herrick often works from multiple photographic referents when executing a single painting. For instance, for his portrait Mayor Edward G. Rendell, Herrick took a photograph of the background fireplace, and then a separate photograph of the mayor, standing. Then in his studio, Herrick observes the first photograph when painting the background, and the second photograph when painting the figure. In this way, Herrick "composits" his painting—the analogy of using Photoshop in digital photography.
Note that there is no guarantee that the lighting conditions in the two photographic referents are the same, nor that Herrick can detect such inconsistencies, nor that he could correct them when executing his painting, even if he detected them. It may be difficult too for the viewer to detect such inconsistencies. Here is where methods from computer vision come into play. The challenge here is that we cannot assume that the light source is a point, such as the candle in Georges de la Tour's Christ in the carpenter's studio.
The basic approach to exposing such conconsistencies in complex lighting environments is to describe the pattern of light intensity on each object with a sum of functions—so-called spherical harmonics. The details are described in our papers, but in brief, these are a set of functions that, added in the proper proportion, can describe any pattern of illumination. If the relative proportions of the lighting components for two objects are the same (or very similar), then it is unlikely one was "compositted" into the scene. But if the relative proportions differ significantly, then they were likely captured under different lighting conditions, and hence at least one was "composited" into the scene.
This technique of lighting analysis may be of use in other works, for instance Renaissance portraits executed in an atelier. If two figures, say, differ significantly in their inferred lighting conditions, then it is likely that they were rendered under different studio conditions, at possibly different times and possibly by different artists.