Hands-On Image Processing with Python

Compact Watershed

As discussed earlier, the watershed algorithm computes watershed catchment basins in an image flooded from given markers, apportioning the pixels into marked basins. The algorithm requires a grayscale gradient image as input (viewing the image as a landscape), where the bright pixels indicate a boundary (forming high peaks) between the regions. From the given markers, this landscape is then flooded until different flood basins meet at the peaks. A different image segment is formed by each distinct basin. As we had with SLIC, there is an additional compactness argument that makes it harder for markers to flood faraway pixels. Higher values of compactness make the watershed regions more regularly shaped. The next code block ...