[ITK-users] testing

[Matt McCormick]

Hi Johannes,

> Thank you! At the moment every test is an executable. Maybe later when the
> tests are getting more and more than I will try to use the test manager.

Sounds like a good plan.


> At the moment I have a problem with the itkTestingComparisonImageFilter. The
> image which is the output of my algorithm is exact the same as the image
> that I use as valid input for this filter which I get from another software.
> But the output of the Comparison filter is max difference -3.40282e+38, the
> min difference +3.40282e+38, total and mean difference are 0, and the
> interesting part is that the number of pixels with difference is 0 (what I
> would expect). But why this min and max values?

These are the default initial values.  Note that the "max" is a very
large negative number and the "min" is a very large positive number.


> In another test I am doing the same where the two images which I compare are
> also the exact same and I get as max difference something in the range of
> 1e-7 and the number of pixels with difference is bigger than 0, but this is
> ok and seems to be the inaccuracy which to expect when using float.

Yes, a tolerance is usually required, especially when working with float pixels.


HTH,
Matt