I have an image that has been segmented into superpixels. Now id like to **merge / grow** similar regions into bigger blobs and I was wondering if anyone can send a link to a paper that discusses the basics of what I am trying to achieve. I have read things like markov / conditional random fields but I don't think they are what I need. Im not trying to label the superpixels. I just want to merge similar ones together.
I was thinking of like a CRF / MRF framework where the nodes are connected to each other at each iteration the number of nodes would get halved or quartered. Meaning to say that the neighbouring superpixels have been grouped. I have been searching for anything related online but have had no luck.
I guess I could just do a simple for loop through the superpixels while computing some similarity function within the 8 neighbours but I was wondering if there is a better framework to achieve what I am trying to do.
↧