In recent years there have been advances in the theory of nonlinear eigenvalue problems related to image processing and computer vision. The formulations of nonlinear transforms, related to one-homogeneous functionals, such as total-variation, has opened way to various applications of image decomposition, face fusion, denoising and more. Theory related to 1-Laplacian eigenvectors on graphs has contributed to better understanding of classification, segmentation and clustering methods. In addition, new numerical methods for solving these hard problems have been proposed. In this two-part minisymposium researchers will present their latest results and discuss future trends in this emerging field.

Introductory words and recent trends in nonlinear spectral processing

Gilboa Guy (Electrical Engineering Department, Technion)

Nonlinear Spectral Image Decomposition and its Application to Segmentation

Zeune Leonie (University of Twente)

Nonlinear spectral methods in machine learning

Matthias Hein (University of Tuebingen)

Spectral total-variation local scale signatures for image manipulation and fusion

Hait Ester (Technion)

Organizers:

Gilboa Guy (Electrical Engineering Department, Technion)

Aujol Jean-Francois (University of Bordeaux)

Keywords:

image segmentation, machine learning, nonlinear optimization, partial differential equation models