You could do color segmentation (see my File Exchange ) to split the image between black and yellow. Then get a binary image of the black parts and pass it to bwlabel() to count the number of black blobs. 1 would be a solid outline, and a broken outline would of course give more parts. You could also call imopen() on the yellow mask to get the interior yellow "square" and then look at the Solidity from the regionprops() if you look at the yellow blob. You could look at the area fraction of black lines and compare it to what it would be if the lines were all connected. There are other things too, but that's a start.