Title :
Solving Large Scale Binary Quadratic Problems: Spectral Methods vs. Semidefinite Programming
Author :
Olsson, Carl ; Eriksson, Anders P. ; Kahl, Fredrik
Author_Institution :
Lund Univ., Lund
Abstract :
In this paper we introduce two new methods for solving binary quadratic problems. While spectral relaxation methods have been the workhorse subroutine for a wide variety of computer vision problems - segmentation, clustering, image restoration to name a few - it has recently been challenged by semidefinite programming (SDP) relaxations. In fact, it can be shown that SDP relaxations produce better lower bounds than spectral relaxations on binary problems with a quadratic objective function. On the other hand, the computational complexity for SDP increases rapidly as the number of decision variables grows making them inapplicable to large scale problems. Our methods combine the merits of both spectral and SDP relaxations -better (lower) bounds than traditional spectral methods and considerably faster execution times than SDP. The first method is based on spectral subgradients and can be applied to large scale SDPs with binary decision variables and the second one is based on the trust region problem. Both algorithms have been applied to several large scale vision problems with good performance.
Keywords :
computer vision; image registration; image segmentation; nonlinear programming; binary decision variables; computational complexity; computer vision; image clustering; image restoration; image segmentation; large scale binary quadratic problems; large scale vision problems; quadratic objective function; semidefinite programming; spectral methods; spectral relaxation methods; trust region problem; Algorithms; Computational complexity; Computer vision; Image restoration; Image segmentation; Large-scale systems; Mathematical programming; Optimization methods; Quadratic programming; Relaxation methods;
Conference_Titel :
Computer Vision and Pattern Recognition, 2007. CVPR '07. IEEE Conference on
Conference_Location :
Minneapolis, MN
Print_ISBN :
1-4244-1179-3
Electronic_ISBN :
1063-6919
DOI :
10.1109/CVPR.2007.383202