Издательство Springer, 2014, -180 pp.International Dagstuhl Seminar, Dagstuhl Castle, Germany, November 20–25, 2011. Revised Selected PapersMost of the leading algorithms in computer vision are based on global optimization methods. Such methods compute the solution of a given problem as minimizer of a suitable cost functional that penalizes deviations from previously made assumptions and integrates them in a global manner, i.e., over the entire image domain. Since this way of modelling is very transparent and intuitive, it is not surprising that such methods have become popular and successful tools to tackle many fundamental problems in computer vision such as, e.g., motion estimation, stereo reconstruction, image restoration, and object segmentation. However, there is also a price to pay when employing global optimization methods. The corresponding cost functionals often lead to optimization problems that are both mathematically challenging and computationally expensive. In order to discuss recent advances and challenges in the design and the solution of global optimization methods, Dagstuhl Seminar 11471 on Efficient Algorithms for Global Optimisation Methods in Computer Vision was held during November 20–25, 2011, at the International Conference and Research Center (IBFI), Schloss Dagstuhl, near Wadern in Germany. The seminar focused on the entire algorithmic development pipeline for global optimization problems in computer vision: modelling, mathematical analysis, numerical solvers, and parallelization. In particular, the goal of the seminar was to bring together researchers from all four fields to analyze and discuss the connections between the different stages of the algorithmic design pipeline. The seminar included researchers from the fields of computer science and mathematics alike. From all submissions, eight high-quality full articles were finally accepted after a strict reviewing process. Each article was reviewed by at least two international experts in the field and only articles with exclusively positive reviews were accepted for publication. The accepted articles reflect the state of the art in the field and focus on recent developments in efficient approaches for continuous optimization and related parallelization aspects on high-end cluster systems.Dense Elastic 3D Shape Matching Fast Regularization of Matrix-Valued Images Recovering Piecewise Smooth Multichannel Images by Minimization of Convex Functionals with Total Generalized Variation Penalty Half-Quadratic Algorithm for lp-lq Problems with Applications to TV-l1 Image Restoration and Compressive Sensing A Fast Algorithm for a Mean Curvature Based Image Denoising Model Using Augmented Lagrangian Method A Smoothing Descent Method for Nonconvex TVq-Models A Fast Continuous Max-Flow Approach to Non-convex Multi-labeling Problems A Geometric Multigrid Solver on Tsubame 2.0
Чтобы скачать этот файл зарегистрируйтесь и/или войдите на сайт используя форму сверху.