Comparison of graph cuts with belief propagation for stereo, using identical MRF parameters
read more
Citations
Computer Vision: Algorithms and Applications
Efficient Belief Propagation for Early Vision
Image Alignment and Stitching: A Tutorial
Convergent tree-reweighted message passing for energy minimization.
Convergent Tree-Reweighted Message Passing for Energy Minimization
References
A taxonomy and evaluation of dense two-frame stereo correspondence algorithms
Fast approximate energy minimization via graph cuts
An experimental comparison of min-cut/max- flow algorithms for energy minimization in vision
Fast approximate energy minimization via graph cuts
Related Papers (5)
Frequently Asked Questions (8)
Q2. How many iterations did the Graph Cuts algorithm take to pass information from one side?
For synchronous updates, the number of iterations must be may need to be as large as the largest dimension of the image in order to pass information from one side of the image to the other.
Q3. How many iterations would it take for information to reach the other side?
For a synchronous update schedule on an image with width W , it would take W iterations for information from one side of the image to reach the other.
Q4. What is the MAP estimate of the whole MRF?
The sum-product algorithm computes the marginal distributions of each node, while the max-product algorithm computes the MAP estimate of the whole MRF.
Q5. What is the way to calculate the probability of the likely labelling of the graph?
When the max-product algorithms converges on a graph with loops, it returns an approximate solution for the most likely labelling of the graph.
Q6. Why is the effect of having large flat regions with sudden jumps caused?
This effect of having large flat regions with sudden jumps is caused because the MAP estimator must assign a single discrete disparity level to each point.
Q7. What is the MAP estimate for the Graph Cuts algorithm?
The authors use the max-product algorithm to find the MAP estimate for comparison with the Graph Cuts algorithm, which also computes the MAP estimate.
Q8. What is the compatibility function of the MRF?
With the compatibility functions defined, the joint probability of the MRF can be written as [1]:P (x1,x2, . . . , xN , y1, y2, . . . , yN ) = ∏(i,j)Ψ(xi, xj) ∏pΦ(xp, yp) (1)where N is the number of nodes, (i, j) represent a pair of neighboring nodes, xn is the variable at location n, and yn is the variable representing the intensity differences.