| |
Methods
104
,
0
abs( )
,
0
x x
x
x x
-
<
=
>=
.
Equation 21: abs
Therefore, if Iij=Gij for an (i,j) in [1,M]x[1,N], the corresponding addend in the sum in
Equation 19 becomes zero for that (i,j). If the ROIs do not match, the addend becomes one.
Summing up over all pixels gives an error value. The maximum error is calculated if no corre-
sponding entries within the two ROI images match. The value of this error (the sum in the
numerator) becomes MN in this case. To normalize the error, the error is divided by MN
(maximal error) so that the normalized error takes on values between zero and one. The qual-
ity Q was defined as (1 - normalized error). Therefore, Q can take on values in the interval [0,
1]. If the quality is maximal (1), the two ROI images are identical. If the quality is 0, no corre-
sponding entries within the two ROI images match.
Calculating the Quality for V1, V2, V3 Separately
The quality can be calculated for V1, V2, and V3 separately. This is done by calculating the
difference matrix after the entries that indicate other visual areas than the current one have
been set to zero in both images (I and G).
5.5 Visualization
It is important to visualize the data in an appropriate way. For example, a suitable color map
can help the user to find boundaries between visual areas easier and more reliably. Figure 49
shows two images that have been created by using different color maps. The same underlying
matrix of phase data from a rotating wedge experiment was used in both cases.
The default color map for wedge data in mrFindBorders uses the same range of non-gray
colors as the default color map in mrLoadRet (red-yellow-green-cyan-blue-magenta-red).
However, in mrFindBorders, this range of non-gray colors is squeezed to the interval [ /2
3 /2]. Phase values outside of this interval correspond to grayish colors. This has two main
|  |
|
| |
|
|
