Paper Filter | Computer Science Papers

@techreport{cs16-001,

title = {Registration Of Nasopharyngoscopic Video With Treatment Planning CT Scans},

author = {Julian Rosenman and Qingyu Zhao and True Price and Marc Niethammer and Ron Alterovitz and Jan-Michael Frahm and Bhishamjit Chera and Stephen Pizer},

url = {https://techreports.cs.unc.edu/papers/16-001.pdf},

year  = {2016},

date = {2016-05-25},

number = {16-001},

abstract = {Purpose: We present a method that allows one to generate data from nasopharyngoscopy videos into a form suitable for use in radiation treatment planning, as well as for review of the data itself as a 3D object.



Methods and Materials:  The accurate transfer of data from endoscopic video frames to the planning CT requires an explicit endoscopic video/CT registration. Our approach to accomplishing this goal is to first reconstruct an accurate 3D model of the anatomy from the 2D video data as seen on multiple video frames. That reconstruction, which we call an endoscopogram, can then be deformably registered to the CT scan.



To generate the endoscopogram we first compute a sparse 3D point cloud by tracking corresponding feature points from one video frame to another using standard structure-from-motion (SfM) techniques adapted to the pharyngeal region. We then use a second technology, shape-from-shading (SFS), to produce high-quality images from each video frame. Uncorrected, these SFS images contain dense information and in many places accurate local curvature, but they lack global spatial fidelity. Using our newly developed techniques, we can correct the SFS images using global information from the sparse SFS-generated point cloud as a constraint. We then register several of these corrected SFS images to each other to generate a 3D image that displays the entire surface area seen in the whole video.



Results: We have successfully computed a spatially accurate endoscopogram (shown below) and have registered it with the planning CT scan with an average accuracy of about 3-5 mms.



Conclusions: The technologies developed by our working group will allow accurate  transfer of data from endoscopy images directly onto the planning CT scan as well as review of the video as a 3D object.

},

keywords = {techreport},

pubstate = {published},

tppubtype = {techreport}

}

2015

Heinly, Jared

Toward Efficient and Robust Large-Scale Structure-from-Motion Systems Technical Report

no. 15-001, 2015.

Abstract | Links | BibTeX | Tags: techreport

@techreport{cs15-001,

title = {Toward Efficient and Robust Large-Scale Structure-from-Motion Systems},

author = {Jared Heinly},

url = {https://techreports.cs.unc.edu/papers/15-001.pdf},

year  = {2015},

date = {2015-12-01},

number = {15-001},

abstract = {The ever-increasing number of images that are uploaded and shared on the Internet has recently been leveraged by computer vision researchers to extract 3D information about the content seen in these images. One key mechanism to extract this information is structure-from-motion, which is the process of recovering the 3D geometry (structure) of a scene via a set of images from different viewpoints (camera motion). However, when dealing with crowdsourced datasets comprised of tens or hundreds of millions of images, the magnitude and diversity of the imagery poses challenges such as robustness, scalability, completeness, and correctness for existing structure-from-motion systems. This dissertation focuses on these challenges and demonstrates practical methods to address the problems of data association and verification within structure-from-motion systems.



Data association within structure-from-motion systems consists of the discovery of pairwise image overlap within the input dataset. In order to perform this discovery, previous systems assumed that information about every image in the input dataset could be stored in memory, which is prohibitive for large-scale photo collections. To address this issue, we propose a novel streaming-based framework for the discovery of related sets of images, and demonstrate our approach on a crowdsourced dataset containing 100 million images from all around the world. Results illustrate that our streaming-based approach does not compromise model completeness, but achieves unprecedented levels of efficiency and scalability.



The verification of individual data associations is difficult to perform during the process of structure-from-motion, as standard methods have limited scope when determining image overlap. Therefore, it is possible for erroneous associations to form, especially when there are symmetric, repetitive, or duplicate structures which can be incorrectly associated with each other. The consequences of these errors are incorrectly placed cameras and scene geometry within the 3D reconstruction. We present two methods that can detect these local inconsistencies and successfully resolve them into a globally consistent 3D model. In our evaluation, we show that our techniques are efficient, are robust to a variety of scenes, and outperform existing approaches.},

keywords = {techreport},

pubstate = {published},

tppubtype = {techreport}

}

The ever-increasing number of images that are uploaded and shared on the Internet has recently been leveraged by computer vision researchers to extract 3D information about the content seen in these images. One key mechanism to extract this information is structure-from-motion, which is the process of recovering the 3D geometry (structure) of a scene via a set of images from different viewpoints (camera motion). However, when dealing with crowdsourced datasets comprised of tens or hundreds of millions of images, the magnitude and diversity of the imagery poses challenges such as robustness, scalability, completeness, and correctness for existing structure-from-motion systems. This dissertation focuses on these challenges and demonstrates practical methods to address the problems of data association and verification within structure-from-motion systems.

Data association within structure-from-motion systems consists of the discovery of pairwise image overlap within the input dataset. In order to perform this discovery, previous systems assumed that information about every image in the input dataset could be stored in memory, which is prohibitive for large-scale photo collections. To address this issue, we propose a novel streaming-based framework for the discovery of related sets of images, and demonstrate our approach on a crowdsourced dataset containing 100 million images from all around the world. Results illustrate that our streaming-based approach does not compromise model completeness, but achieves unprecedented levels of efficiency and scalability.

The verification of individual data associations is difficult to perform during the process of structure-from-motion, as standard methods have limited scope when determining image overlap. Therefore, it is possible for erroneous associations to form, especially when there are symmetric, repetitive, or duplicate structures which can be incorrectly associated with each other. The consequences of these errors are incorrectly placed cameras and scene geometry within the 3D reconstruction. We present two methods that can detect these local inconsistencies and successfully resolve them into a globally consistent 3D model. In our evaluation, we show that our techniques are efficient, are robust to a variety of scenes, and outperform existing approaches.

Cheng, Wei

Toward Robust Group-wise Eqtl Mapping Via Integrating Multi-domain Heterogeneous Data Technical Report

no. 15-002, 2015.

2020

2018

2017

2016

2015

2014

2013

2012

2011

2010