Article text
Anyone combining images needs to have a basic
understanding of perspective. In the early part
of my professional life I was a TV cameraman,
one of my favourite kinds of show was what was
known in the business as a CSO (or Chromakey
depending on which company you worked for) epic.
A show where all the sets and locations were
either paintings or photographs - think “Sky
Captain and the World of Tommorow” but made
(mostly) without computers. One of the main
things a cameraman had to do on these shows was
to get the actors to convincingly move around
their “virtual” world. This required a
reasonable knowledge of how perspective works so
when the background painting had totally
unrealistic perspective or the photo background
had been drastically cropped you could make the
actors look as much as possible like they were
in the set.

If you are combining images and you want the
combination to be seamless this is the
fundamental idea to work from:


Thus if you have a photo of a person taken with
a camera 15 feet away and 3 foot off the ground,
the shot of the background must also be taken
with the camera 3 foot off the ground (as long
as the ground is level of course) with the
position you want to place the person 15 feet
away from the camera. No amount of fiddling in
Photoshop can correct this if it isn’t right!
Consideration must also be made for where the
person is going to be in the frame. If the shot
of the person has them in the centre of the
frame (i.e. the cameras optical axis goes
through the centre of the person) then the
optical centre of the background picture must
be where the person is going to be positioned.
If you wish to have the person off centre either
shoot them in the correct position in the frame
in the first place or either the foreground or
the background needs to have its perspective
altered which can either be done using a shift
lens at the taking stage or with the transform
tools in Photoshop.

In practical terms this is what I do. When
shooting the people for my images I shoot them
against a plain coloured backdrop (usually blue)
with a grid of yellow dots on the floor. This
helps to assess the tilt of the camera, if the
camera is vertical following the line of dots
back the point where they meet will be at the
centre of the photo , if the camera is tipped
down the dots will meet at a point higher than
the centre and lower than the centre if the
camera is tipped up. I mostly use miniature
backgrounds made to 1/6th scale, this allows me
to use a highly posable action figure as a stand
in. I normally shoot the photo of the person
first then I build the miniature and position
the action figure where I want the person to be
in the final image. Then I frame up the camera
so I get the same shot of the action figure as
the shot of the person, using the same focal
length lens (picked up from the exif data from
the original shot). If the shot is tilted I use
a sheet of card with a scaled down version of
the grid of dots on it so I can make them match.
The camera is then reframed using the zoom and
panning and tilting it around the front nodal
point of the lens (the front nodal point of a
lens is normally slightly behind the front
element). If you just pan or tilt normally the
optical centre of the picture will move, the
amount ,however, is less significant if the set
is full size. I normally try to include a
reference horizontal and vertical to make
transforming the picture in Photoshop easier.