Hi there,
I have a doubt about the question: "When using the ST min-cut algorithm to segment an image:"
One of the correct answers says, "The algorithm operates on very sparse graphs." But aren't we supposed to consider each pixel as a node, hence the entire image forms a dense graph, right? I can only think of making the graph sparse by assuming the nodes to be superpixels acquired by running a 5D K-Means beforehand.
I really appreciate any help you can provide!
Hi Vishal,
In the graph defined in Slide 59 of Segmentation, edges only connect neighboring pixels.
Or maybe the misunderstanding here is that the sparsity of a graph refers to it's number of edges, not nodes. Here, |E| is roughly equal to 4 |V|, which is low compared to a dense graph of ~|V|² edges.
Hope that helps!
Corentin
In the graph defined in Slide 59 of Segmentation, edges only connect neighboring pixels.
Or maybe the misunderstanding here is that the sparsity of a graph refers to it's number of edges, not nodes. Here, |E| is roughly equal to 4 |V|, which is low compared to a dense graph of ~|V|² edges.
Hope that helps!
Corentin
Aha, that makes sense!
Thank you!
Thank you!
