Discrete Visual Perception

Abstract

Abstract-Computational vision and biomedical image have made tremendous progress of the past decade. This is mostly due the development of efficient learning and inference algorithms which allow better, faster and richer modeling of visual perception tasks. Graph-based representations are among the most prominent tools to address such perception through the casting of perception as a graph optimization problem. In this paper, we briefly introduce the interest of such representations, discuss their strength and limitations and present their application to address a variety of problems in computer vision and biomedical image analysis.

Graph-Based Slice-to-Volume Deformable Registration

“…Low, mid and highlevel vision tasks can be addressed within this framework. To this end, a visual perception task is addressed by specifying a task-specific parametric model, associating it to the available observations (images) through an objective function and optimizing the model parameters given both, the objective and the observations [36].…”

(Hyper)-Graphs Inference through Convex Relaxations and Move Making Algorithms: Contributions and Applications in Artificial Vision

“…The introduction of efficient inference algorithms inspired from the networks community, like for example the max flow/min cut principle at late nineties that is a special case of the duality theorem for linear programs as well their efficient implementations towards taking advantage of image like graphs Boykov et al [1998] or message passing methods Pearl [1998] that are based on the calculation of the marginal for a given node given the states of the other nodes have re-introduced graphical models in the field of computer vision. During the past two decades we have witnessed a tremendous progress both on their use to address visual perception tasks Wang et al [2013], Kappes et al [2015], Paragios and Komodakis [2014], Szeliski et al [2008], Blake et al [2011], Komodakis and Tziritas [2007a] as well as it concerns their inference. This tutorial aims to provide an overview of the state of the art methods in the field for inference as well as the most recent advances in that direction using move making algorithms and convex relations.…”