Bug in ImageLinePruneMerge.pruneSimilar

[Erhan Erdemir]

I think there is a small bug in the function

ImageLinePruneMerge.pruneSimilar

The relevant snippet is as below:

// now just see if they are very close
float distA = Distance2D_F32.distance(a, b.a);
float distB = Distance2D_F32.distance(a, b.b);

if (distA > toleranceDist && distB > toleranceDist) {
continue;
}

distA = Distance2D_F32.distance(b, b.a);
distB = Distance2D_F32.distance(b, b.b);

if (distA > toleranceDist && distB > toleranceDist) {
continue;
}

It first finds and checks the distance between line a and end points of line b.

Then it finds and checks the distance between line b and end points of line b.

I think the second one is wrong. It should find and check the distance between line a (not b) and end points of line b.

Thanks

[Peter A]

Hi Erhan,

I'll take a closer look at this soon. Just wanted to say something since it's been over a week. Thanks for reporting this!

Cheers,

- Peter

--
You received this message because you are subscribed to the Google Groups "BoofCV" group.
To unsubscribe from this group and stop receiving emails from it, send an email to boofcv+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/boofcv/a26e5d5d-9117-4007-b913-fb6deb495f6bn%40googlegroups.com.

--

"Now, now my good man, this is no time for making enemies."    — Voltaire (1694-1778), on his deathbed in response to a priest asking that he renounce Satan.

[Peter A]

Yep definitely a bug. However I'm not sure if it makes a big difference or not. Added a unit test (there was none before) and had trouble causing the old code to generate a mistake. After thinking about it I believe in the nominal situation the two sets of distances will be identical. It might not be the case if they intersect different parts of the image border.

https://github.com/lessthanoptimal/BoofCV/pull/386

This will be merged once it's been verified to not cause any performance issues.