~~
Speaker: Ilai Bistritz,
M.Sc. student under the supervision of Prof. Amir Leshem and Prof. Hagit Messer - Yaron

Wednesday, June 29th, 2016 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Game Theoretic Tools for Resource Allocation in Wireless Communication Networks

Abstract

We consider the problem of distributed channel allocation in large networks under the frequency-selective interference channel. Performance is measured by the weighted sum of achievable rates, when treating interference as noise. We propose two non-cooperative random game formulations for this problem, and analyze their Nash equilibria probabilistically and asymptotically in the number of players. While the first game is natural, we show that it suffers from asymptotically poor performance, when interference is sufficiently strong. The second game uses a slight modification of the utility function of the first. We show that all its Nash equilibria have asymptotically close to optimal performance, for any interference regime and for a broad class of fading distributions. In order to exploit these results algorithmically we propose a modified Fictitious Play algorithm that can be implemented distributedly without any communication between the players or any central entity.

~~
Speaker: Naftali Zon
M.Sc. student under the supervision of Prof. Nahum Kiryati

Wednesday, June 22nd, 2016 at 15:00
Room 011, Kitot Bldg., Faculty of Engineering

Unified Functional Framework for Restoration of Image Sequences Degraded by Atmospheric Turbulence

Atmospheric turbulence between the scene and the camera degrades image quality. Diverse algorithms estimate the latent image from an image sequence of a static scene. The flow of these algorithms reflects various concepts, but their theoretical foundations are not sufficiently clear. Recent algorithms have yielded dramatic progress, yet many aspects of the problem remain challenging and hard to understand. The goal of this thesis is to analyze recent turbulence restoration algorithms and to suggest a unified variational framework to address the turbulence restoration problem. A unified functional quantifies the association between a given image sequence and a candidate latent image restoration. Minimizing the functional using advanced numerical methods leads to a general algorithmic flow. I show that various known algorithms can be derived as special cases of the general approach. Furthermore, I show that building-blocks used in turbulence recovery algorithms, such as optical flow estimation and blind deblurring, are called for by the general model. The main contribution of this thesis is the establishment of a unified theoretical framework for the restoration of turbulence-degraded images. It leads to novel turbulence recovery algorithms as well as to better understanding of known ones. Along the way we develop a state of the art method for general-purpose blind deconvolution and a novel method for optical flow estimation.

You are invited to attend a lecture

Snapshot Spectral Imaging Using a Digital Camera with an Optical Diffuser and Compressed Sensing Algorithms

By

Jonathan Hauser

M.Sc. student of

Professor Michael Golub and Professor Menachem Nathan

Physical Electronics Department, Tel Aviv University

Spectral Imaging (SI) systems are designed to capture the spatial and spectral information of an object or a scene, which is referred to generally as "data cube" or "spectral cube". There is a strong link between the spectrum, physical state and components of matter. Accordingly, SI systems are practical in various fields such as medicine, agriculture, food inspection, military and homeland security, as well as in various academic research fields related to physics, chemistry, biology, geography, engineering and more.

Many of the SI systems that exist in the industry are based on scanning techniques, are quite complicated and require long spectral cube acquisition time. Snapshot Spectral Imaging (SSI) systems aim to capture the spectral cube with a single snapshot, thus eliminating the need of scanning the spatial and/or spectral space of the scene. Known SSI systems are generally complex and expensive in terms of their electrical, optical and mechanical implementation.

The aim of this study is to realize a working prototype of a compact SSI system based on a standard monochromatic digital camera, a single thin optical diffuser and a computational unit. The diffuser is a transparent phase element, which is designed to fulfill Compressed Sensing (CS) mathematical conditions that enable the spectral cube reconstruction from a single "shot".

The achieved results show, for the first time, feasibility proof for successful restoration of a full spectral cube, from a snapshot taken with a monochromatic image sensor in a relatively simple CS-SSI system.

Monday, June 13, 2016, at 09:00

Room 011, Kitot Building

~~
Speaker: Chen Harel
M.Sc. student under the supervision of Prof. Daniel Cohen-Or and Prof. Shai Avidan

Wednesday, June 15th, 2016 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering

Inference in Perturbation Models for Enhanced Matching

Matching meaningful correspondences is a challenging and important problem in computer vision. In particular, bipartite graph matching is a useful tool in many visual applications. In this work, we introduce the notion of perturbations for enhanced matching, and present the power of perturbation matching in the context of shape matching of image contours.
Perturbation models are high dimensional probability models that measure the stability of max-prediction to random shifts of the computed scores. The probability of a given matching is the volume of scores for which this matching is the highest scoring matching.
We demonstrate the effectiveness of our approach in pair of images with background clutter, showing that inference with perturbation models is more robust than finding the highest scoring matching. We also show that for perturbation matching it is favorable to use logistic perturbation models rather than Gumbel perturbation models, proving that inference using logistic perturbation models is more efficient.