[Insight-users] Comparing images only with same size + tolerance?

[Luis Ibanez]

Hi Reinhold,

Thanks for letting us know of the success
of your experiment. It is quite healty to
verify registrations using an alternative
image metric.

In your case, it may also be a good idea to
use the NormalizeCorrelation as a metric for
further refining the results of the registration
performed with MeanSquares. In this case you
will need a third metric in order to evaluate
the quality of the result (e.g. PatternIntensity
could be a good option for this third metric).


Regards,


   Luis


-------------------------
Reinhold Füreder wrote:

 > Dear Luis,

....

 >> What is the normal range of intensities in
 >> your images ?
 >> 0:255 ?
 >> or are they having pixel types float ?
 >>
 >> The mean differences that you get may be
 >> actually very high is the dynamic range
 >> of the pixels is  [0:1].
 >
 >
 >
 > Alright. It is [0..1] and so I misinterpreted the result of the
 > filter...
 >
 >
 >> NormalizedCorrelation is a good metric for
 >> evaluation since it has a very pointy center.
 >> That is, the value decreases rapidly when
 >> the images are not aligned correctly.
 >
 >
 >
 > With the NormalizedCorrelation image metric I eventually get the desired
 > results; typical values (for the different flippings) are:
 >
 >   0: -0.43559
 >   1: -0.273582
 >   2: -0.689396
 >   3: -0.223274
 > Comparison DOES comply!!!
 >
 > As the optimal value of this metric is minus one (cf. software guide)
 > the result seems to be quite convincing./?
 >
 >
 > Thanks for your enlightenment Luis,
 >     Reinhold
 >