Application of Active Contours on Metallurgical Microstructures
- PDF / 1,402,408 Bytes
- 12 Pages / 593.972 x 792 pts Page_size
- 106 Downloads / 212 Views
INTRODUCTION
PERIODIC sectioning and reconstructing threedimensional (3D) metallurgical microstructures using a computer first was employed in 1994 to obtain virtual images of pearlite and the discontinuous transformation for subsequent analyses.[1,2] Reconstructed 3D models eliminate false interpretations and some errors introduced by statistical analysis of two-dimensional (2D) data. Image segmentation (e.g., extracting a phase of interest from the matrix), a process that is either done by manually outlining the microstructure or by using a lowlevel threshold-based edge detector such as Canny, is a typical intermediate step in the reconstruction of 3D models. However, these segmentation techniques often require extensive supervision (user interaction) or a normalizing step (e.g., histogram equalization) for datasets with variable contrast. As a result, the numerous images often required for 3D models renders these techniques impractical. Active contouring (AC) is a segmentation approach developed in the late 1980s for meteorology and has been applied in several fields including medicine and bioengineering. Although AC is an edge-based segmentation technique, it is less susceptible to contrast differences, which makes it ideal to use across datasets with varying contrast within regions of interest (ROI) without employing normalizing algorithms. More importantly, edge linking and spur removal (examples of these image outlining defects are ALEKS Y.M. ONTMAN, Graduate Student, and GARY J. SHIFLET, WG Reynold Professor, are with the Department of Materials Science, University of Virginia, Charlottesville, VA 22904. Contact e-mail: [email protected]. Manuscript submitted March 28, 2009. Article published online May 22, 2010 2236—VOLUME 41A, SEPTEMBER 2010
shown in Figure 2(d)) procedures often associated with Canny edge detectors (or the similar Roberts and Sobel edge detectors), are not needed for AC. The current article is the first in a series that will demonstrate methods to automate completely the 3D reconstruction process by outlining, registering, and volumizing (3D reconstruction) a series of digital micrographs once the images are secured. The current article introduces an automated outlining procedure, which will be followed by the automatic registration of a series of images using artificial neural networks. Volumizing the microstructure then can be accomplished using commercial software.
II.
INTRODUCTION TO ACTIVE CONTOURS
An active contour is a curve that evolves from a predefined initial position toward the boundaries of a desired object. Active contours can be divided into the following general classes: geometric and parametric. Each class has its own implementation that makes its behavior unique. Both classes find use in a broad range of applications.[3] After a brief introduction to AC types, metallurgical examples are presented, illustrating their individual usefulness in segmenting microstructures. A. Introduction to Parametric Active Contour (PAC) Parametric active contour (PAC) first was introduced by
Data Loading...
