Location

Brunswick, ME

Start Date

1-1-1992 12:00 AM

Description

In NDE images, one often comes across overlapping or connected regions of interest. Segmentation of these regions becomes one of the preprocessing stages before any kind of image analysis. A secondary electron image showing silicon carbide particles dispersed in aluminum metal matrix composites is given in Figure 1. To study the microstructural characteristics of these composites, segmentation of the various particles is essential. Segmentation of these particles using conditional skeleton was about forty percent successful [1]. Three different approaches with varying degrees of complexity and accuracy are discussed in this paper. All three techniques are based on the principles of mathematical morphology [2]. The first step in all three techniques is to generate a marker or seed for each particle in the image. Once the seeds are generated, the techniques differ in the manner in which the seeds are grown. All the techniques work on binary images. Experimental results with microstructural statistics for each technique are presented in the paper. Some of the advanced morphological tools used are defined in the next section

Book Title

Review of Progress in Quantitative Nondestructive Evaluation

Volume

11A

Chapter

Chapter 3: Interpretive Signal Processing and Image Reconstruction

Section

Signal Processing

Pages

887-894

DOI

10.1007/978-1-4615-3344-3_114

Language

en

File Format

application/pdf

Share

COinS
 
Jan 1st, 12:00 AM

Study of Microstructural Characteristics Using Mathematical Morphology

Brunswick, ME

In NDE images, one often comes across overlapping or connected regions of interest. Segmentation of these regions becomes one of the preprocessing stages before any kind of image analysis. A secondary electron image showing silicon carbide particles dispersed in aluminum metal matrix composites is given in Figure 1. To study the microstructural characteristics of these composites, segmentation of the various particles is essential. Segmentation of these particles using conditional skeleton was about forty percent successful [1]. Three different approaches with varying degrees of complexity and accuracy are discussed in this paper. All three techniques are based on the principles of mathematical morphology [2]. The first step in all three techniques is to generate a marker or seed for each particle in the image. Once the seeds are generated, the techniques differ in the manner in which the seeds are grown. All the techniques work on binary images. Experimental results with microstructural statistics for each technique are presented in the paper. Some of the advanced morphological tools used are defined in the next section