Multi-Projector Content Preservation with Linear Filters

Abstract

Using aligned overlapping image projectors provides several ad-
vantages when compared to a single projector: increased bright-
ness, additional redundancy, and increased pixel density within
a region of the screen. Aligning content between projectors is
achieved by applying space transformation operations to the de-
sired output. The transformation operations often degrade the qual-
ity of the original image due to sampling and quantization. The
transformation applied for a given projector is typically done in iso-
lation of all other content-projector transformations. However, it is
possible to warp the images with prior knowledge of each other
such that they utilize the increase in effective pixel density. This
allows for an increase in the perceptual quality of the resulting
stacked content. This paper presents a novel method of increas-
ing the perceptual quality within multi-projector configurations. A
machine learning approach is used to train a linear filtering based
model that conditions the individual projected images on each other