Template matching based on normalized-cross-correlation (NCC) algorithm uses the minimization of squared Euclidean distance
at position
to find best matching between a reference object 
(with dimension
)
and a region 
in an image at position
.
The Euclidean distance is minimized, when the linear cross correlation coefficient
between the reference object 
and image region
is maximized. To account for intensity
variations in the image and make the correlation coefficient invariant to pixel intensities, the correlation between the
difference of object
to the mean value
and
image region
to the mean value
is calculated.
According to [Bur06] the so called normalized-cross correlation coefficient can then be expressed as:
with
The result of 
is between -1 and 1.
The higher the accordance between reference image 
and image region 
is, the higher is result
of 
.
The position 
in image
with highest value of
,
is the location where the reference object 
is found.
Prev
