Following the RTcam project we decided to add colour and other non-geometric attributes into our camera. These turned out to be easily included in the same mathematical framework, which is already very simple.
The ARTcam allows effects such a colour aberration, depth of field, star-burst filters, etc. It allows users to specify a lens by drawing. It can be calibrated to real optical devices like mirrors and lenses - here coffee pots and glass tumblers, which allows for parametric environment matting - which unlike non-parameteric methods does not suffer from a “blind spot”. Any colour change can be calibrated for too, this colour constancy facility is built directly into the ARTcam. Typical publications include:
Top left, an original photograph. Top right, a depth of field filter. Bottom left, areal perspective (colours are bluer
and less contrast with depth). Bottom right, depth of field and areal perspective combined.
Left, an original photograph. Middle colour misprinting, right colour aberration - both simulated using ARTcams.
Left is an original photograph, right is a modified version. The cluster of photographs are correctly
reflected in the coffee pot, the colour of which has been mapped from a christmas bauble.
C. Li and P. Hall, “Colour constancy based on model selection”, Proc. British Machine Vision Conference, 122-131, 2007.
C. Li and P. Hall, “Environment matting into photographs using coloured RTcams”, Proc. Conference on Visual Media Production, 1-8, 2008.
C. Li and P. Hall, “ARTcams: attributed rational tensor cameras”, Proc. Computational Aesthetics, 73-91, 2009.