S.-L. Liu, R. R. Martin, F. C. Langbein, P. L. Rosin
Computer-Aided Design and Applications, 4(5):565-583, 2007.
ISSN 16864360.
[Published] [Preprint] [CiteSeer]
Segmentation of geometric reliefs from a textured background has various applications in reverse engineering. We consider two approaches to solve this problem. The first classifies parts of a surface mesh as relief or background, and then uses a snake which moves inwards towards the desired relief boundary, which is coarsely located using an energy based on the classification. The second approach initially smoothes the surface to eliminate the background texture, and locates the snake at the relief boundary using an energy based on the step between the background and the relief. Both snakes start at simple user-drawn contours, and are driven towards the relief boundaries by the snake energy functional. In both cases, the snake has different evolution phases with different energy terms, to initially rapidly drive the snake towards the relief boundary, and to later accurately match it. To describe geometric textures, we analyze surface differential properties, and integral and statistical quantities based upon them, computed at multiple scales taken over local neighborhoods, following similar ideas from image texture processing. For classification, we use a support vector machine together with sequential forward floating search for feature selection. A straightforward Laplacian method is used for smoothing. We use example scanned models to demonstrate that both approaches are useful, but are suitable for different types of model.
@ARTICLE{Liu2007,
author = {Shenglan Liu and Ralph R. Martin and Frank C.
Langbein and Paul L. Rosin},
title = {Segmenting Geometric Reliefs from Textured
Background Surfaces},
journal = {Computer-Aided Design and Applications},
year = 2007,
volume = 4,
pages = {565-583},
number = 5,
issn = 16864360,
url = {http://www.langbein.org/research/surfaces/reliefs/liu2007/},
abstract = {Segmentation of geometric reliefs from a textured
background has various applications in reverse
engineering. We consider two approaches to solve
this problem. The first classifies parts of a
surface mesh as relief or background, and then uses
a snake which moves inwards towards the desired
relief boundary, which is coarsely located using an
energy based on the classification. The second
approach initially smoothes the surface to eliminate
the background texture, and locates the snake at the
relief boundary using an energy based on the step
between the background and the relief. Both snakes
start at simple user-drawn contours, and are driven
towards the relief boundaries by the snake energy
functional. In both cases, the snake has different
evolution phases with different energy terms, to
initially rapidly drive the snake towards the relief
boundary, and to later accurately match it. To
describe geometric textures, we analyze surface
differential properties, and integral and
statistical quantities based upon them, computed at
multiple scales taken over local neighborhoods,
following similar ideas from image texture
processing. For classification, we use a support
vector machine together with sequential forward
floating search for feature selection. A
straightforward Laplacian method is used for
smoothing. We use example scanned models to
demonstrate that both approaches are useful, but are
suitable for different types of model. },
}
Segmenting Geometric Reliefs from Textured Background Surfaces,http://www.langbein.org/index.php?id=161 by Frank C Langbein [ 6/December/2008, 19:24].
Copyright © 1995-2010, Frank C Langbein.
Except where otherwise stated, this web site, Ex Tenebris Scientia, by Frank C Langbein is licensed under a Creative Commons Attribution-Non-Commercial-Share Alike 2.0 UK: England & Wales License.
