Can Eigenvalue Centrality Detect Blobs in Images?

[NotASmurf]

Hey all, I just read up on the principle of centrality, where

"Think of a "network" as an NxN matrix, which has information about how N people (or N pages or N countries..) are connected to each other.

Adjacency Matrix is an NxN matrix, let's say it looks something like this. People who aren't connected to each other have A[j] = 0, people with weak relationships have A[j] = 0.1, people with medium ties have A[j]=0.4,
people with strong ties have A[j] = 0.6
1---2----3----4
1---0.0 0.1 0.4 0.6
2---0.1 0.0 0.4 0.0
3---0.4 0.4 0.0 0.1
4---0.6 0.0 0.1 0.0

1 and 2 are weakly connected, 1 and 3 have medium ties, 1 and 4 have strong ties. This is just a quick example to give you a quick idea. These matrices may not always be symmetric either.

People with "higher eigenvector centrality" are people who are better connected to each other. This takes into account, not just how many people the person knows, but also whom the person knows.'

Is it feasible to use this to locate blobs in images using their color relations to each other? I need a feature detector/ blob extractor that doesn't use corners so I saw this. Any advice appreciated.

 

[mfb]

How do you plan to use it? The eigenvectors don't care about order, so the concept of "being next to each other" (which is relevant if you look for features) doesn't translate well to the matrix problem.

 

[kroni]

What is the final goal ? Send a picture, i will see if there is no simpler method.
The method you are explaining need to compute eigen vector of a sparse matrix of size PixelNumber * PixelNumber. You can use préconjugate gradient but it's not easy.