Skip to main content
main-content

Über dieses Buch

Since the early days of computers, machine learning and automatic programming have attracted researchers in computer science and related fields, particularly pattern recognition and automatic control theory. Most of the learning concepts in machine perception have been inspired by pattern recognition approaches that rely on statistical techniques. These statistical techniques have applicability in limited recognition tasks. Automatic programming in perception systems has generally been limited to interfaces that allow easy specification of the task using natural language. Clearly, machine learning and automatic programming can make percep­ tion systems powerful and easy to use. Vogt's book addresses both these tasks in the context of machine vision. He uses morphological operations to implement his approach which was developed for solving the figure-ground problem in images. His system selects the correct se­ quence of operators to accept or reject pixels for fmding objects in an image. The sequence of operators is selected after a user specifies what the correct objects are. On the surface it may appear that the problem solved by the system is not very interesting, however, the contribution ofVogt' s work should not be judged by the images that the system can segment. Its real contribution is in demonstrat­ ing, possibly for'the frrst time, that automatic programming is possible in computer vision systems. The selection of morphological operators demonstrates that to implement an automatic programming-based approach, operators whose behavior is clearly defined in the image space are required.

Inhaltsverzeichnis

Frontmatter

1. Introduction

Abstract
The project described here was motivated by months of my sitting in front of an image processing system, trying different parameters and operations on a series of similar images, in the hope of developing one algorithm to work for all cases. (The problem was to try to locate the blood vessels in digital coronary angiograms). I remarked that I was spending an enormous amount of time computing partial results for each example image, testing the applicability of operations to subproblems, and looking for ideal parameter ranges over all of the examples considered together. It seemed to me that much of this work could be automated, simply by developing high level routines to do testing and parameter searches on a whole series of examples at once, rather than having me test each one ‘by hand’. Going further, it seemed that it should be possible to formalize the applicability of operations to a particular problem, and beyond that to formalize the entire search process for an algorithm. That is what this book is about, and it essentially arose out of my own experience and frustrations at performing these same steps manually for a period of two years while at Thomson-CGR in France, and subsequently at Machine Vision International (MVI) and at the Environmental Research Institute of Michigan (ERIM) in the United States.
Robert C. Vogt

2. Review of Mathematical Morphology

Abstract
We begin our study of the automatic generation of set recognition algorithms by giving a brief review of the essential elements of mathematical morphology. Though we cannot hope to cover all of it in such a short space, I will describe the main principles of the theory and give a general overview of the major classes of operations, while concentrating on those which made up the focus of this project. The purpose of this review is to provide the mathematical underpinnings for what is to follow. Some of the material in this chapter, particularly that in Sections 2.3, 2.5, 2.7, 2.10 and 2.11, is treated in greater depth in Serra [1982].
Robert C. Vogt

3. Theory of Automatic Set Recognition

Abstract
In this chapter I want to describe, both formally and informally, the process of algorithm development for set recognition tasks, including what are the essential components of this process and what are some of some of the techniques and strategies that can be used to improve results. The treatment given here will take the form of an overview; for certain topics additional detail will be found in Chapters 4 and 5, which describe the implementation and its current capabilities.
Robert C. Vogt

4. REM System Implementation

Abstract
The chapter which follows provides a complete description of the REM program, which was implemented to serve as a demonstration of the concepts described in Chapter 3. Not everything discussed in Chapters 2 and 3 has been implemented, but enough has been to serve as a ‘proof of principle’ of the ideas presented there.
Robert C. Vogt

5. Results

Abstract
This chapter presents the results of the implementation described in Chapter 4, divided into four parts. Section 1 gives a summary of the system’s capabilities; Section 2 illustrates examples of the range of different problems solved by the system as part of the program test and evaluation effort. Section 3 describes one complete example run, from start to finish, and Section 4 discusses the issues of efficiency, i. e., how fast is the system compared to a human or compared to an unintelligent machine search?
Robert C. Vogt

6. Conclusion

Abstract
There are several elements of this work which I regard as primary accomplishments or contributions, and which deserve special mention here. Some of these are related to the central theme of the work, the automatic generation of image processing algorithms, while others are related more to the domain of operators I chose to work with, that of mathematical morphology. Additionally there are some lesser contributions, related to the implementation effort, which I want to list at the end of this section.
Robert C. Vogt

Backmatter

Weitere Informationen