Minimal surfaces based object segmentation

Abstract. Fitting parameterized 3D shape and general reflectance models to 2D image data is challenging due to the high dimensionality of the problem. The proposed method combines the capabilities of classical and photometric stereo, allowing for accurate reconstruction of both textured and non-textured surfaces. In particular, we present a variational method implemented as a PDE-driven surface evolution interleaved with reflectance estimation. The surface is represented on an adaptive mesh allowing topological change. To provide the input data, we have designed a capture setup that simultaneously acquires both viewpoint and light variation while minimizing self-shadowing. Our capture method is feasible for real-world application as it requires a moderate amount of input data and processing time. In experiments, models of people and everyday objects were captured from a few dozen images taken with a consumer digital camera. The capture process recovers a photo-consistent model of spatially varying Lambertian and specular reflectance and a highly accurate geometry.

show abstract

“…This flow is accurate for a g that is only a function of surface position x. Similar PDE's were used by [18,16] but with different g functions.…”

Section: Surface Evolutionmentioning

confidence: 94%

Variational Shape and Reflectance Estimation Under Changing Light and Viewpoints

Birkbeck

Cobzaş

Sturm

et al. 2006

Computer Vision – ECCV 2006

View full text Add to dashboard Cite

show abstract

“…Nevertheless, this approach has been extended to surfaces in a 3D image by extracting a minimal surface laying on two given curves [3]. The advantage of this method is that it does not suffer from local minima problems, as would other active surface methods like [4,5].…”

Section: Introductionmentioning

confidence: 99%

From a Single Point to a Surface Patch by Growing Minimal Paths

Benmansour

Cohen

2009

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. We introduce a novel implicit approach for surface patch segmentation in 3D images starting from a single point. Since the boundary surface of an object is locally homeomorphic to a disc, we know that the boundary of a small neighboring domain intersects the surface of interest on a single closed curve. Similarly to active surfaces, we use a cost potential which penalizes image regions of low interest. First, Using a front propagation approach from the source point chosen by the user, one can see that the closed curve corresponds to valley line of the arrival time from the source point. Next, we use an implicit 3D segmentation method. It assumes that the object boundary contains two known constraining curves. In our case, the first curve is reduced to a point and the other one is automatically detected by our approach. A partial differential equation is introduced and its solution is used for segmentation. The zero level set of this solution contains valley line and the source point as well as the set of minimal paths joining them. We present a fast implementation which has been successfully applied to 3D biomedical and synthetic images.

show abstract

“…One of the earliest work on volumetric lung nodule segmentation reported in literature was by Kawata et al [195,196] which adopted the geodesic AC approach by [225]. El Baz and Farag et al [197,198] have adopted the energy minimization approach by Kass et al [223] with a prior appearance model by Markov random field (MRF) and a current appearance model by a bi-modal linear combination of discrete Gaussians.…”

Section: Deformable Model (Dm) Represents a Class Of Segmentation mentioning

confidence: 99%

Computational methods for the analysis of functional 4D-CT chest images.

Soliman¹

View full text Add to dashboard Cite

DEDICATIONThis dissertation is dedicated to my father's soul.iii ACKNOWLEDGMENTS In the name of Allah the most merciful, the most compassionate. All deepest thanks are due to Almighty Allah for the uncountable gifts given to me. Medical imaging is an important emerging technology that has been intensively used in the last few decades for disease diagnosis and monitoring as well as for the assessment of treatment effectiveness. Medical images provide a very large amount of valuable information that is too huge to be exploited by radiologists and physicians. Therefore, the design of computer-aided diagnostic (CAD) system, which can be used as an assistive tool for the medical community, is of a great importance. This dissertation deals with the development of a complete CAD system for lung cancer patients, which remains the leading cause of cancer-related death in the USA. In 2014, there were approximately 224,210 new cases of lung cancer and 159,260 related deaths. The process begins with the detection of lung cancer which is detected through the diagnosis of lung nodules (a manifestation of lung cancer). These nodules are approximately spherical regions of primarily high density tissue that are visible in computed tomography (CT) images of the lung. The treatment of these lung cancer nodules is complex, nearly 70% of lung cancer patients require radiation therapy as part of their treatment. Radiation-induced lung injury is a limiting toxicity that may decrease cure rates and increase morbidity and mortality treatment. By finding ways to accurately detect, at early stage, and hence prevent lung injury, it will have significant positive consequences for lung cancer patients.v The ultimate goal of this dissertation is to develop a clinically usable CAD system that can improve the sensitivity and specificity of early detection of radiation-induced lung injury based on the hypotheses that radiated lung tissues may get affected and suffer decrease of their functionality as a side effect of radiation therapy treatment. These hypotheses have been validated by demonstrating that automatic segmentation of the lung regions and registration of consecutive respiratory phases to estimate their elasticity, ventilation, and texture features to provide discriminatory descriptors that can be used for early detection of radiation-induced lung injury. The proposed methodologies will lead to novel indexes for distinguishing normal/healthy and injured lung tissues in clinical decisionmaking. To achieve this goal, a CAD system for accurate detection of radiation-induced lung injury that requires three basic components has been developed. These components are the lung fields segmentation, lung registration, and features extraction and tissue classification.This dissertation starts with an exploration of the available medical imaging modalities to present the importance of medical imaging in today's clinical applications. Secondly, the methodologies, challenges, and limitations of recent CAD systems for lung cancer detection are covered...

show abstract

Minimal surfaces based object segmentation

Cited by 178 publications

References 34 publications

Variational Shape and Reflectance Estimation Under Changing Light and Viewpoints

Variational Shape and Reflectance Estimation Under Changing Light and Viewpoints

From a Single Point to a Surface Patch by Growing Minimal Paths

Computational methods for the analysis of functional 4D-CT chest images.

Contact Info

Product

Resources

About