Information Technology Reference
In-Depth Information
Figure 3.58:
The basic components of a phase correlator.
The result is a set of frequency components which all have the same amplitude, but have phases corresponding to
the difference between two fields. These coefficients form the input to an inverse transform.
Figure 3.59
(a) shows
what happens. If the two fields are the same, there are no phase differences between the two, and so all the
frequency components are added with zero degree phase to produce a single peak in the centre of the inverse
transform. If, however, there was motion between the two fields, such as a pan, all the components will have phase
differences, and this results in a peak shown in
Figure 3.59
(b) which is displaced from the centre of the inverse
transform by the distance moved. Phase correlation thus actually measures the movement between fields.
Figure 3.59:
(a) The peak in the inverse transform is central for no motion. (b) In the case of motion the peak shifts
by the distance moved. (c) If there are several motions, each one results in a peak.
In the case where the line of video in question intersects objects moving at different speeds,
Figure 3.59
(
c) shows
that the inverse transform would contain one peak corresponding to the distance moved by each object.
Whilst this explanation has used one dimension for simplicity, in practice the entire process is two-dimensional. A
two-dimensional Fourier transform of each field is computed, the phases are subtracted, and an inverse two-
dimensional transform is computed, the output of which is a flat plane out of which three-dimensional peaks rise.
This is known as a correlation surface.
Figure 3.60
(a) shows some examples of a correlation surface. At (a) there has been no motion between fields and
so there is a single central peak. At (b) there has been a pan and the peak moves across the surface. At (c) the
camera has been depressed and the peak moves upwards.
