Research Papers

Creeping Contours: A Multilabel Image Segmentation Method for Extracting Boundary Surfaces of Parts in Volumetric Images

[+] Author and Article Information
Haitham Shammaa, Hiromasa Suzuki, Yutaka Ohtake

Research Center for Advanced Science and Technology, The University of Tokyo, Komaba 4-6-1, Meguro, Tokyo 153-8904, Japan

The results presented in this paper were generated on a standard workstation equipped with a Xeon 2.5 GHz CPU and 8.00 Gbyte of main memory.

J. Comput. Inf. Sci. Eng 11(1), 011007 (Mar 31, 2011) (9 pages) doi:10.1115/1.3569830 History: Received May 24, 2010; Revised September 16, 2010; Published March 31, 2011; Online March 31, 2011

In this work, we introduce a method named creeping contours for image segmentation into component parts for the purpose of extracting the boundary surfaces of these parts. Creeping contours are contours that expand following a speed function defined by the gradient and curvature at contour points, starting from an initial contour position defined either manually or automatically. Contours in the image creep simultaneously at different speeds, while labels are assigned to contour pixels by the defined creeping condition. We also demonstrate the effectiveness of the proposed method by segmenting 2D grayscale images and 3D volumetric computed tomography images of mechanical parts into multiple segments and generating the boundary surfaces of these parts.

Copyright © 2011 by American Society of Mechanical Engineers
Your Session has timed out. Please sign back in to continue.



Grahic Jump Location
Figure 1

(a) Assembly of motorbike pedal. (b) Reconstructed boundary surfaces of different materials with the object’s CT volumetric data; missing parts or unwanted outliers are framed.

Grahic Jump Location
Figure 2

Example of an iteration in growing one creeping contour using a pseudo-2D-image grid. Pixels of R are shown in dark color, surrounded by creeping contour pixels (framed). (a) Values for approximate curvature (κ(p)) calculated over 3×3 kernels in an eight-neighborhood system (smaller values represent concave regions, while higher values represent convex ones); (b) pseudovalues for normalized gradient (‖γI‖); (c) Fcur(p)—the result of adding closure pixels (light color) satisfying κ(p)≤2 from the creeping contour to the segmented region (the region takes on a more compact shape); (d) Fgrad(p)—the result of adding closure pixels (light color) satisfying ‖γI‖≤4 from the creeping contour to the segmented region; (e) F(p)—the final pixels (light color) added to the inside region, which are those shared between (c) and (d).

Grahic Jump Location
Figure 3

Illustrative model of four creeping contours in a 2D space at four points in time: (a) initial seed regions, (b) start of creeping contours, (c) after several creeping contour iterations, and (d) at the end of the image segmentation process

Grahic Jump Location
Figure 4

Segmentation of a multimaterial 2D CT image into three segments when the image includes local spray noise: (a) original image, (b) initial seeds, (c) intermediate stage of creeping contours at iteration 30, and (d) segmentation result. λgrad=0.1 and λgrad=0.1 for the three materials.

Grahic Jump Location
Figure 5

Segmentation of a 2D CT image of a tooth cross-section into three segments: (a) original image, (b) initial seeds, (c) restricted region-growing result, and (d) segmentation result

Grahic Jump Location
Figure 6

Segmentation of a 2D MRI cross-section of the human brain in the transverse (upper row) and sagittal (bottom row) planes: (a) original image with initial seeds, (b) segmentation result, and (c) segmentation result overlaid on the original image

Grahic Jump Location
Figure 7

Different settings for segmentation of a 2D CT cross-section of two parts made of the same material and coming into contact with a line: (a) original image with initial seeds, (b) segmentation result, and (c) segmentation result overlaid on the original image

Grahic Jump Location
Figure 8

Segmentation results for natural gray images from the Berkeley Segmentation Dataset: (a) result of human segmentation from the Berkeley Segmentation Dataset overlaid on the original image, (b) original image with initial seeds, and (c) segmentation result overlaid on the original image

Grahic Jump Location
Figure 9

Extraction of boundary surfaces from two blocks (the smaller one made of aluminum and the larger one made of rubber) using (a) only the threshold values of the materials in the CT data and (b) the creeping isocontour method

Grahic Jump Location
Figure 10

Extraction of boundary surfaces from a vibration damper containing steel and rubber parts at different speed functions of creeping isocontours. The calculation time is 1308 s.

Grahic Jump Location
Figure 11

Extraction of boundary surfaces from motorbike engine parts containing steel and aluminum materials. The surface of the steel contains 115,426 triangles, and that of the aluminum has 1,028,682 triangles. The segmentation time is 163 s. (a) Half cross-section; (b) boundary mesh surfaces of the whole object.

Grahic Jump Location
Figure 12

Extraction of boundary surfaces from motorbike pedal parts containing steel, aluminum, plastic, and rubber materials: (a) segmentation using graph cut; the rings show outliers generated by the global characteristic of the graph cut algorithm; (b) segmentation using creeping isocontours; the rings show thin parts that could be reconstructed using creeping contours but not graph cut

Grahic Jump Location
Figure 13

Extraction of boundary surfaces from thin sheets containing steel and aluminum with different speed functions for the creeping isocontours




Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In