Fan-beam reconstruction in computer tomography from full and partial projection data

Miscellaneous · Fri Jan 01 00:00:00 EST 1988

OSTI ID:6472811

Peng, H

The problems of image reconstruction in this thesis appear in medical tomography, nondestructive testing, astrophysics, electromicroscopy and other fields. The problem is to image the internal structure of the object of interest from the shadows (raysums) at different angles. Fan-beam reconstruction is required when the probing source is a point source. Fan-beam geometry also has the advantage of fast scanning in computer tomography. The conventional reconstruction algorithm (Filter/Convolution Back projection) requires O(N{sup 3}) computations to reconstruct a 2-D image. Investigated are an alternative way to do fan-beam reconstruction which uses O(N{sup 2}log N) computations through the use of the direct Fourier reconstruction method. Partial-data image reconstruction occurs when the collection of full data is impossible due to an internal opacity, an exterior obstruction, imaging a time varying event, or reducing the dosage or scanning time. Other prior knowledge is incorporated of the image function to compensate for the effect of missing data. The explicit minimum distance solution is derived for the incomplete data problem in a one step implementation of the iterative algebraic reconstruction technique (ART). This solution is a projector and can be used with projector of other prior knowledge constraint sets in the framework of the method of Projection Onto Convex Sets (POCS). The POCS algorithm solution has all the known properties of the object and is consistent with all the data. We also formulate a new set which utilizes prior information based on the similarity of the unknown object to some known reference. The full algorithm for partial-data reconstruction using POCS is implemented. Simulations demonstrate the efficacy of the method. The shape of an opacity from the raysum data is directly estimated and used with a performance analysis in terms of system parameter.

• Research Organization: Rensselaer Polytechnic Inst., Troy, NY (USA)
• OSTI ID: 6472811
• Resource Relation: Other Information: Thesis (Ph. D.)
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE
COMPUTER GRAPHICS
IMAGE PROCESSING
ALGORITHMS
COMPUTER ARCHITECTURE
COMPUTERIZED SIMULATION
COMPUTERIZED TOMOGRAPHY
FOURIER TRANSFORMATION
INFORMATION SYSTEMS
MATHEMATICAL MODELS
RESEARCH PROGRAMS
DIAGNOSTIC TECHNIQUES
INTEGRAL TRANSFORMATIONS
MATHEMATICAL LOGIC
PROCESSING
SIMULATION
TOMOGRAPHY
TRANSFORMATIONS
990200* - Mathematics & Computers
990300 - Information Handling