Connectivity preserving transformation of digital images: Theory and applications.
Item
-
Title
-
Connectivity preserving transformation of digital images: Theory and applications.
-
Identifier
-
AAI9510688
-
identifier
-
9510688
-
Creator
-
Ma, Cherng-Min.
-
Contributor
-
Adviser: T. Yung Kong
-
Date
-
1994
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Computer Science
-
Abstract
-
Thinning is a preprocessing operation of pattern recognition that erodes an object layer by layer until only a unit width skeleton is left. A thinning algorithm must "preserve connectivity," that is, for every possible input image, its output image has the same connectivity structure. Many thinning algorithms for 2D images have been proposed. But this is not the case for 3D images.;Our first objective is to design two 3D fully parallel thinning algorithms--one for generating skeletons as "medial faces" and the other one for "medial lines." A multimedia team at M.I.T. is interested in the medial-face algorithm for analyzing actions of baseball players in 3D images (length {dollar}\times{dollar} width {dollar}\times{dollar} time). The medial-line algorithm is applied to 3D medical images at the University of Iowa. The images (length {dollar}\times{dollar} width {dollar}\times{dollar} depth) show bronchial trees in human lungs. The resulting images after the application of our algorithm can be used eventually ta save lives.;Since there are infinitely many images, it is difficult to show that a thinning algorithm preserves connectivity for all input images. Ronse proposed a solution to this problem for 2D images. Our second objective is to establish connectivity preservation tests for 3D thinning algorithms such that whether a 3D thinning algorithm does or does not preserve connectivity can be determined efficiently and systematically.;Since both Ronse's 2D results and our 3D results only need to check a finite number of configurations, we can use computers to give such proofs automatically. Kong et al. established such a 2D program using a small look-up table for testing the connectivity preservation of any thinning algorithm. Our third objective is to establish computerized tests using 3D look-up tables to verify the connectivity preservation of 3D thinning algorithms.;A thinning algorithm is a set of deleting templates. We use a look-up table to test the connectivity preservation of each template. A complete 3D look-up table is about 318 GBytes which is not feasible. Thus, our last objective is to establish a memory efficient algorithm for establishing such a 3D look-up table of a feasible size.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
