How to determine if a set of pixels are connected to each other?

[softmessager]

I have a binarized image of a greyscale image .
How to detect if there is a crack on that image ? ...
If its a Horizontal line, we can find it out by a for loop enclosed in an outer for loop (for column and row)
But cracks can come Vertically or diagonally or not a straight line at all..
Please help...
I am able to view the cracks (pixels) in my image (set of arranged white pizels) ..

A Brief overview of my work.
I am working on vision inspection of a bottle that is finished filling and capped I want to determine whether the bottle has a crack.I am able to get the white pizels in my image But dont know how to determine whether its a crack. Also there might be other white pixels for bubbles or suspended particles.
How to distinguish them from a crack.

My platform is VC6 with OpenCv

Expecting help
Softmessager

[philkr]

I think because a crack can be of so different looks, that the best solution would be a neural network. You could then give it pictures of bottles with cracks to learn how they look.
But independent of what you do, your problem is to reduce the image information. Maybe this "trace contour" algorithm of some graphics programs will be of some help. You might find something with google maybe.