Rob Atkinson ha scritto:
> One could create a index page of output artefacts, for a quick visual 
> scan. The human brain is pretty good. If this index page included static 
> "expected" samples it would be pretty quick to eyeball.

Hum... one of the major features we want out of a build is to
be headless and automated, so a visual comparison  would have to
be dumped on some file for human evaluation, which is not going
to be automatic.

Yet, the visual comparison could be included in the release producers
making sure we don't have busted releases like 2.4.3.

> If one also filtered out pixel-identical images it might be a relatively 
> small and simple sanity check to do.

> any "picture closeness" measures could be used in conjunction, and even 
> a pixel-diff mask created by someone in future - this would be an easy 
> place for (SoC?) people to play and add capability.

Soc is already gone, we'd have to wait for next year.
As a quick measure of distance, we could do the following. For each
pixel, extract the R G B A components, make a difference, component
by component, against the expected value, and sum up the absolute
values of it. If the total sum on all the pixels exceeds a certain
soil, fail the test, and present a visual exemplar of the comparison.

On failures that do not represent a real issue, we can just relax the
soil accordingly.

What other people think of such approach?
Cheers
Andrea

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
Geotools-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geotools-devel

Reply via email to