Intersect() Problem when none of the pixels overlap

When I used intersect() to find the intersection between A (red) and B (blue), I failed to get any because the pixels don't overlap. What can I do to estimate the intersection point?

[Ashish Uthama]

How would you define (for a computer) an intersection? For example, what if they just 'touch'? (Image 1 and 2 represent A and B)

1 0 0 2 
0 1 2 0
0 1 2 0
1 0 0 2

[Hg]

In my case, the A (red) is a line so I can thicken or extend it to get an intersection.

Answer 1

[Image Analyst]

There is no one pixel where the overlap occurs. If you'll accept any of those 4 pixel locations as an overlap, then perhaps if you dilated, ANDed, then called bwulterode. Something like (untested)

intImage = imdilate(bw1, true(3)) & imdilate(bw2, true(3));
intPoints = bwulterode(intImage);

bwulterode() has some connectivity options, so you might play around with those.

[Hg]

Thanks. I'll use it for future reference. In my case right now, since A (red) is a line so I will just thicken or extend it to get an intersection.

[Image Analyst]

That's exactly what the first line of code does. It thickens the lines. Then I AND it to find the intersection. It sounds like you think your method is different somehow, but I think they're the same. What I do in my second line of code, to prevent getting an intersection point that is thicker, is to call bwulterode() which shrinks everything down to a single point.

Another thing you could try is to OR the two binary images together and then use bwmorph() to get branchpoints:

intPoints = bwmorph(bw1 | bw2, 'Branchpoints');

[Hg]

Got it. It is exactly what I need. Thanks for the insight!