Shape Estimation Using Graph Cut

This dissertation is mainly concerned with shape estimating of an object from its two-dimensional (2D) images. In this dissertation, various shape estimation problems are formulated as global optimization problems which are solved using graph cut. Particularly, an approach called graph cuts based active contours (GCBAC) is proposed, and its applications to 2D object segmentation and three-dimensional (3D) object modeling are discussed. In the application of object segmentation, the presented approach represents the 2D image as an edge capacitated graph and iteratively applies multisource multisink minimum cut on the graph where an object is embedded to extract the shape of the object. In the application of modeling a 3D object from its 2D images captured from different viewpoints, a node capacitated graph is constructed from the input images and therefore the approach is also called iterative node cut. Graph cuts based algorithms are also presented in this dissertation for interactive object selection, object contour tracking, and omnifocus image generation. In addition, this dissertation presents two non-graph-cut approaches for 3D shape estimation before applying GCBAC approach to 3D shape estimation.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical Engineering

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Xu, Ning

Contributors dc:contributor

• Ahuja, Narendra

• Computer Science

Language dc:language

eng

Identifier

(MiAaPQ)AAI3182426

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/80912