Image processing methods in materials science

Carsten Kruse Olsson

The topic of this thesis is image processing methods in material science. The topic is approached through a number of cases and several examples.

The whole process from data acquisition to results is described for three cases in the first part of the thesis. A considerate amount of the description is concerned with the image processing.

The image processing methods which were used in these cases and in several other examples are presented in the second part of the thesis.

Chapter 5 is concerned with linear filters. The chapter covers methods to reduce noise, enhance edges and lines, to detects 'anormalities' in an image, and methods to recognize textures. A discrete version of the Canny edge detection filter is developed in section 5.3.1. A spin-off of this is a smoothing filter with some interesting properties regarding the gradient in noisy images. This smoothing filter is used in a new approach to line detection, section 5.4. Image models are used to predict pixel values in section 5.5 and it is shown how different statistical tests on the residuals can be used to find 'interesting' areas in the image.

The chapter on Mathematical Morphology contains an introduction to grids as well as the normal morphological operations. New and efficient implementations are, however, described for several interesting grey level structuring elements. The properties of these structuring elements are described and they are used both here and in other examples in the rest of the thesis.

Chapter 6 is concerned with operations which involve morphological definifions or functions which are used as a part of different morphological operations. The Labeling algorithm is central for many of the operations in this chapter such as the Hole Fill, the Double Threshold, and the Local Maximum algorithms. The Watershed algorithm is shown to be a very useful way to segment grey level images, and some effective ways to filter the segmentation made by the watershed are also described.

Finally, automatic (un-supervised) thresholding is discussed in chapter 8 and existing methods are developed further.

IMSOR Ph.D Thesis 64, 1993

Last modified August 18, 1996
Finn Kuno Christensen

Back to IMM Homepage