Files in this item
|(no description provided)|
|Title:||Integrated 3D Analysis and Analysis Guided Synthesis|
|Doctoral Committee Chair(s):||Ahuja, Narendra|
|Department / Program:||Electrical Engineering|
|Degree Granting Institution:||University of Illinois at Urbana-Champaign|
|Subject(s):||Engineering, Electronics and Electrical
|Abstract:||This thesis has two goals. First, we are concerned with the reliable estimation of the motion and structure of rigid objects from a variety of cues in monocular images under perspective projection. Second, we synthesize images that depict the estimated motion and structure.
The thesis is divided into three parts. In the first part, we present various methods for the estimation of motion and structure from different, single cues and different numbers of frames: (i) estimation of the orientation of planar surfaces in a single image using the texture gradient and vanishing line, (ii) robust nonlinear estimation from point correspondences in two images, (iii) linear estimation from line correspondences in two images when all lines are on the same plane, (iv) linear estimation from region correspondences in two frames, and (v) robust nonlinear estimation from point correspondences in multiple frames under the assumption of constant rotational velocity.
In the second part, we are concerned with the integrated use of multiple image cues: (i) We present an algorithm which matches and simultaneously segments the multiple cues (points, regions and lines) in two frames. (ii) We present an approach to recovering three-dimensional motion and structure parameters from multiple cues in an image sequence. For concreteness, we focus on flight image sequences. The approach presented allows the use of cues such as regions, points, flow, lines, texture gradient and vanishing line. The integration of the information from these diverse cues is performed using optimization. A sequential batch method is used for reliable estimation.
In the third part, we are concerned with analysis-guided synthesis: (i) We introduce the notion that the cues that contribute the most to three-dimensional interpretation are also the ones that would yield the most realistic synthesis, thus suggesting an approach to analysis-guided compression. For synthesis, real and/or artificial attributes are shown as a monocular or a binocular sequence. (ii) We present a method for the analysis-guided recognition and synthesis of a runway in flight images.
All of the algorithms presented are followed by experimental results for simulated and real images to show the feasibility of the algorithms.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 1993.
|Date Available in IDEALS:||2014-12-16|
This item appears in the following Collection(s)
Dissertations and Theses - Electrical and Computer Engineering
Dissertations and Theses in Electrical and Computer Engineering
Graduate Dissertations and Theses at Illinois
Graduate Theses and Dissertations at Illinois