Y. Boykov and V. Kolmogorov. A new algorithm for energy 32 minimization with discontinuities. In Intl. Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, pages 205--220, 2001.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....DP, 3 SO) Our implementations of Graph cuts (4) and Bayesian diffusion (5) are several orders of magnitude slower. The authors implementations of the graph cut methods (10 and 11) however, are much faster than our implementation. This is due to the new max flow code by Boykov and Kolmorogov [21], which is available at www.cs.cornell.edu People vnk software.html. In summary, if efficiency is an issue, a simple shiftablewindow method is a good choice. In particular, method 14 by Hirschmuller [53] is among the fastest and produces very good results. New implementations of graph cut methods ....

Y. Boykov and V. Kolmogorov. A new algorithm for energy 32 minimization with discontinuities. In Intl. Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, pages 205--220, 2001.

....labeling of minimum cost. We remark that if the distance function d is not a metric, then determining whether a graph can be colored by k colors is a special case of the labeling problem. A prototypical application of the metric labeling problem is the image restoration problem in computer vision [3, 4, 5]. In the image restoration problem, the goal is to take an image corrupted by noise and restore it to its true version. The image consists of pixels and these are the objects in the classification problem. Each pixel has a discretized intensity value associated with it that is possibly corrupted ....

Y. Boykov, O. Veksler, and R. Zabih. A new algorithm for energy minimization with disconitnuities. In International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, 1999.

....approximation preserving. In [4; 6; 5; 21] they also developed some flow based local search heuristics for labeling problems using the uniform metric. Their methods can be thought of as clever extensions of the multiway cut Isolation Heuristic of Dahlhaus et al. 10] For one of their heuristics [6], they prove that any local optimum is a 2 approximation. In fact, it is possible to adapt their proof to show that the local improvement heuristic provides a 2 approximation in polynomial time. In [4; 21] they show that the heuristics also perform very well in practice. They also consider ....

....f of the vertices of the graph G. It then makes a simple local move whose goal is to decrease the cost of the labeling. If making the move causes it to reach a new labeling whose cost is lower than the cost of the current labeling, it moves to the new labeling and continues. Boykov et al. [6] use the following local improvement step. Consider a label i. In a single local improvement step they allow relabeling any subset of the vertices with the label i, and they call this local move an i expansion move. They prove that if the distance function on the labels is a metric, then the ....

[Article contains additional citation context not shown here]

Yuri Boykov, Olga Veksler, and Ramin Zabih. A new algorithm for energy minimization with discontinuities. In International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, 1999.

....labeling of minimum cost. We remark that if the distance function d is not a metric, then determining whether a graph can be colored by k colors is a special case of the labeling problem. A prototypical application of the metric labeling problem is the image restoration problem in computer vision [3, 4, 5]. In the image restoration problem, the goal is to take an image corrupted by noise and restore it to its true version. The image consists of pixels and these are the objects in the classification problem. Each pixel has a discretized intensity value associated with it that is possibly corrupted ....

Y. Boykov, O. Veksler, and R. Zabih. A new algorithm for energy minimization with disconitnuities. In International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, 1999.

....In addition for the piecewise constant prior we develop an algorithm that nds a local minimum in yet another interesting move space. All the methods we develop use graph cuts as a powerful optimization technique. Some of the work presented in this thesis was previously published in [9] [11], and [10] 16 1.7 Outline In chapter 2 we present some technical tools, justify minimization of the energy functions in this thesis using MAP estimation of certain Markov random elds, and give some examples of vision problems. Chapter 3 presents an algorithm for an everywhere smooth prior. ....

Y. Boykov, O. Veksler, and R. Zabih. A new algorithm for energy minimization with discontinuities. In International Workshop on Energy Minimization Methods in Computer Vision and Pattern Recognition, 1999. 87 88

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC