The pioneering start of image processing at NBS led to many pioneering applications of image processing. The earliest was the creation of the field of quantitative metallography. In the early 1960's, G. A. Moore of the NBS Metallurgy Division, having heard of the scanner, asked R.A. Kirsch whether he might use it for stereology applications. It was common practice, at that time to determine the size distribution of particles in metallurgical photomicrographs by drawing random chords through such images. The length of the intersections with the particles could be tabulated to give a statistical estimate of particle size. Moore proceeded to create the field of quantitative metallography as it is now known (13). Most commercial machines used today in quantitative metallography can be traced back to this fundamental work.
In the NBS Standard Reference Materials program, precise measurements of standard materials were necessary. In later work, Interrante and Hicho used scanned images to calibrate such materials.(14)
Much of the research at NBS was concerned with measurement science as noted above. Another kind of measurement more related to image pattern recognition was for the recognition of fingerprints. In an attempt to automate fingerprint processing for the FBI, Wegstein, Moore and Rafferty developed recognition algorithms (15).
A more orthodox class of image measurements were made by E.C.Teague and his colleagues to support the study of surface topography using interferometry (16). In some of this work, laser interferograms were scanned and measurements made on them produced precise descriptions of surface roughness.
An important source of energy conservation is the use of recycled fibers in the manufacture of paper. But, before such fibers can be used, they must be characterized in terms of their dimensional properties. E.L.Graminski and R.A. Kirsch showed that by scanning photomicrographs of paper fibers, measurements could be made on them that would predict their properties when used to make recycled paper (17).
The image processing field has spawned research in various kinds of vision systems. These include vision systems that robots could use in navigating the world. A group at NBS led by J.Albus has been exploring robot vision systems since 1979 (18). Much of this work is concerned with robots for use in manufacturing.
In the early 1960's the National Institutes of Health (NIH) began considering the automation of cytology, the measurement of cell structure in biological microscopy. Only the simplest kinds of cell analyses were available at the time. The NBS image processing experience was useful in designing an ambitious program in automated cytological analysis that resulted in a large laboratory program at the NIH (19). This included a general plan for cell analysis as well as the development of specific methods for analysis of cell images.
As image processing began to be seen as useful in areas outside the physical sciences and outside metrology narrowly construed, it became apparent to Kirsch that assigning structure to images was a generalization of the kinds of numerical measurement that NBS had been doing for many years. The success of architects (20) in describing the structure of design objects led Kirsch to investigate the use of image processing in the fine arts (21). The tools that architects had used were applied to the structural analysis of painting style (22). One consequence of this structural representation was an economical representation of paintings (23). Kirsch has recently applied these tools to the study of archaeological objects like rock engravings (24).
<< previous next >>
Exhibit Home | Introduction | SEAC Contributions | Evangelism | Testing | Early Image Processing |
Consequences | Development of Image
Processing | New
Processing Tools | Conclusion | References