Method for segmenting images using superpixels and entropy rate clustering
An image is segmented into superpixels by constructing a graph with vertices connected by edges, wherein each vertex corresponds to a pixel in the image, and each edge is associated with a weight indicating a similarity of the corresponding pixels, A subset of edges in the graph are selected to segment the graph into subgraphs, wherein the selecting maximizes an objective function based on an entropy rate and a balancing term. The edges with maximum gains are added to the graph until a number of subgraphs is equal to some threshold.
1. A method for segmenting an image into superpixels, comprising the steps of:
constructing a graph with vertices connected by edges, wherein each vertex corresponds to a pixel in the image, and each edge is associated with a weight indicating a similarity of the corresponding pixels;
selecting a subset of edges in the graph to segment the graph into subgraphs, wherein the selecting maximizes an objective function, wherein the objective function is submodular and includes an entropy rate of a random walk on the graph to produce homogeneous superpixels and a balancing term including an entropy component of distribution of membership of segments maximizing a number of segments of similar sizes and a number of connected subgraphs component minimizing a number of segments; and
adding the edge with a maximum gain to the graph until a number of subgraphs is equal to a threshold, and repeating the selecting and adding steps if the number of subgraphs is below the threshold, wherein the steps are performed in a processor.
2. The method of claim 1 , wherein each subgraph includes homogeneous and similar sized superpixels.
3. The method of claim 1 , wherein the entropy rate is submodular and monotonically increasing.
4. The method of claim 1 , wherein the balancing term is submodular and monotonically increasing.
5. The method of claim 1 , wherein a constraint on a number of subgraphs in a cycle-free graph is a matroid.
6. The method of claim 1 , objective function is maximized using a greedy process subject to a constraint.
7. The method of claim 1 , wherein optimality is guaranteed to be ½ of a global minimum of the objective function.
8. The method of claim 7 , wherein the greedy process is implemented using a heap structure.
9. The method of claim 1 , wherein the segmentation is achieved hierarchically.
10. The method of claim 9 , wherein the hierarchy forms multiple segmentations of the image simultaneously.
11. The method of claim 1 , wherein a balancing parameter of the balancing term is tuned automatically.
12. The method of claim 1 , wherein a balancing parameter of the balancing term is modified by user to customize segmentation.
13. The method of claim 1 , wherein the segmentation is performed interactively with user supervision.
14. The method of claim 1 , wherein general clustering problems in non-image domains are solved.