. What is the main difference between flood fill and connected-component labeling?
Flood fill needs one seed per region, while cv2.connectedComponentsWithStats scans the whole image and labels every blob at once.
. In the classic stack-based flood_fill_stack algorithm, what happens when a pixel is popped from the stack?
Popped pixels that are background or already filled are simply skipped; otherwise they're marked filled and their neighbors are added to the frontier.
. What does cv2.connectedComponentsWithStats label as component 0?
Label 0 is always the background, so the number of actual blobs is num_labels - 1.
. What extra information does cv2.connectedComponentsWithStats provide alongside the labels, at almost no extra computational cost?
The lesson notes these stats come essentially free as part of the same labeling pass.