Oral-History:William F. Schreiber

About William F. Schreiber

Born in New York City in 1925, William Schreiber attended Columbia University on a full scholarship. With his studies interrupted by the outbreak of WWII, he joined the Navy where, shortly before the end of the war, he was invited to volunteer for the atomic bomb tests at Ataloucock (but declined). After the war, he completed his Bachelor’s degree in electrical engineering at Columbia, followed by a Master’s degree in electrical engineering under the GI bill. After two years working for Sylvania, Schreiber returned to school, obtaining a Ph.D. from Harvard in Applied Physics.

Schreiber went to work for Technicolor corporation where he was contracted to build a television-based simulator that could be used for printing color film. Following his years at Technicolor, he moved to MIT as an associate professor where he remained until he retired. Honored as a Bernard Gordon professor, he attempted to introduce practicality to MIT’s science-based curriculum. Simultaneous to his tenure at MIT, Schreiber was involved in the research and development of mechanical, CRT, and laser scanners. Through ECRM (his own company, founded to develop optical character recognition) he designed, among other products, a computer-based color printing system for newspaper page layout. His current interest lies in writing columns concerning economics and questions of social policy.

About the Interview

WILLIAM F. SCHREIBER:An Interview Conducted by Frederik Nebeker, IEEE History Center, 6 October 1998

Interview #345 for the IEEE History Center, The Institute of Electrical and Electronics Engineering, Inc.

Copyright Statement

This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.

Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.org. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

It is recommended that this oral history be cited as follows:

William Schreiber, an oral history conducted in 1998 by Frederik Nebeker, IEEE History Center, Piscataway, NJ, USA.

Interview

Interview: William Schreiber

Interviewer: Rik Nebeker

Date: 6 October 1998

Place: Chicago, Illinois

Education and naval service

Nebeker:

I’d like to start by asking you to tell us where and where you were born and a little bit about your family.

Schreiber:

I was born in New York City in 1925. I went to the New York City public schools. I got a full scholarship to go to Columbia. I would have tried for Cooper Union if I hadn’t gotten a scholarship. My parents were in no position to pay for a college education.

Nebeker:

What did your father do?

Schreiber:

He was a salesman for electric light fixtures. My mother was a bookkeeper from time to time. Neither one of them was educated much. My mother’s ambition for me was to be a high school teacher. Things changed when the WWII started in 1941, shortly after I had started at Columbia.

Nebeker:

You were about 16 or 17, around about?

Schreiber:

I was 16. My brother went into the Air Force (he is four years older than I am) and I joined a Navy training program after I did one year at Columbia under my Pulitzer scholarship. After six more terms, which occurred in two calendar years, I graduated, went into midshipman’s school, and was commissioned an Ensign in the Navy. The war was still on.

Nebeker:

What year was that you were commissioned?

Schreiber:

It must have been late 1944 or early 1945. After finishing midshipman’s school and being commissioned, I went to a training school in Georgia. Later, I was stationed in Boston for a short time, and then, just before the war was over, I was assigned to a ship, a small wooden subchaser, located in Hawaii. On this little ship, I then sailed back to the west coast, acting as navigator. I was then invited to volunteer for the atomic bomb tests at Enwietok, but I declined.

Nebeker:

As one of the observing ships?

Schreiber:

Yes. I declined because I was worried I was about to lose my girlfriend and I wanted to get home as soon as possible. Well, I got home as soon as possible, but I lost her anyway. In retrospect, it was one of the best things that ever happened to me. I went back to Columbia where I got a master’s degree under the GI Bill.

Graduate studies; Sylvania employment

Nebeker:

Was that in physics, your master’s?

Schreiber:

It is in Electrical Engineering. Shortly after graduating, I started working at Sylvania.

Nebeker:

May I ask who were your influential teachers?

Schreiber:

Well Ragazzini was one of my teachers who was well known in his field. I did go to him for some professional advice because I had been offered a job with the CIA (or whatever the agency was known at that time, I don’t really remember) and he advised me against it. He said I should get into engineering and get some experience. It was good advice.

Nebeker:

You got a job with Sylvania you said?

Schreiber:

Yes, after I got my Masters, I worked at Sylvania in Bayside, New York, under a very nice fellow, Mike Liefer, a one-time high-school teacher. My formal job was designing telemetering circuitry. However, I spent much of my time studying Shannon, because Shannon’s original paper on information theory came out at that time, and I was totally fascinated. Mostly as a result of reading Shannon, I decided I needed to go back to school and get a Ph.D. I thought I wanted a Ph.D. in physics. I applied everywhere. I was accepted at Harvard except Harvard said they did not think I really wanted to be a physicist from my background. They urged me to enter their Engineering Science and Applied Physics program. That was very, very good advice. I spent four years at Harvard and learned a lot.

Nebeker:

Before talking about that period, I wondered if you could tell me if the Sylvania experience was formative in any way.

Schreiber:

Well, it was interesting because they had some government contracts to build electronics equipment, telemetering equipment, and I did get involved in that, so I got a fair amount of experience designing circuits. I became aware of the necessity for doing these things very carefully.

Nebeker:

And I can see how Shannon’s paper might be something that you would see as very important if you have been working on the telemetering.

Schreiber:

Yes, because I thought it was a way to get at the heart of what a process was all about. The fact you could define information mathematically and evaluate different kinds of modulation systems according to their efficiency and capability for transmitting information was something that appealed to me very much. The mathematics that was involved in Shannon’s early papers was well within my grasp. I didn’t have any trouble with it at all because the mathematics courses at Columbia were quite good, even if the engineering work was quite out of date. I elected to do a PhD thesis related to what Shannon had written about.

I had a very nice advisor, at least he was nice to me, Harry Mimno. Harry was British -- English or maybe Scotch. At some point, I met Harry’s mother. What amazed me was that she spoke perfect English. Up until then I had thought that all mothers came from the old country and spoke broken English. Of my own four grandparents, I knew only one, with whom my mother, and therefore I, spent a lot of time before I went to school. She was a Yiddish speaker. She spoke some English but she preferred to speak Yiddish. That was the secret language between my mother and her mother so that I shouldn’t know what they were talking about.

Nebeker:

You probably picked up.

Schreiber:

Of course I picked up a fair amount, but not to the point of speaking it.

At that time in the PhD program at Harvard you needed to be qualified in two foreign languages. My first choice was French, which I had studied in high school and was very easy for me, and I picked German for the second language because, at that time, German was quite useful to American scientists. (We used a German language text in one of my physics courses.) I gave myself a crash course, using a marvelous book I picked up on teaching German to people who didn’t know it for purposes like this. I think the guy’s name was Pollard from University of Texas. Anyway, much of that German sounded so familiar to me, you know, because the particular Yiddish that my grandmother spoke was low German. At that time, there was a song that was sung by Judy Garland called “Bei mir bist du schoen” (By me, you are beautiful) and I understood it. I mean I knew the words. It was kind of nice. Actually later on I became extremely anti-German because of the war and because of what the Germans did to the Jews. It took me decades to get over this prejudice. I was never an observant Jew, but I felt myself to be a Jew nevertheless. I still do. Anyway, getting back to the main thread.

Nebeker:

Now what was Mimno’s area?

Schreiber:

Ionospheric propagation. That was his area. But he was perfectly willing to supervise my thesis; in fact, what he did was leave me alone. He was a very shy person not given to interfering with what I was doing. Of course I kept him informed. He thought this was perfectly okay. I thought it was perfectly okay. What I was doing was measuring the second-order probability distribution of television signals. I was into compression from these very early days, and Shannon tells you the amount of compression you can have depends on the entropy. So what I did was to measure the entropy of TV signals for the first time.

Nebeker:

Now why exactly were you into compression?

Schreiber:

Well, I was always interested in pictures. …

Nebeker:

The Sylvania work didn’t have to do with pictures, right? Or did it?

Schreiber:

No, what they were doing was telemetering. But when the Shannon paper came out in 1948, my boss at Sylvania was very happy to have me study Shannon. With my supervisor I even wrote a sort of popular paper on that subject that appeared in a series called Advances in Electronics.

Nebeker:

This was your Columbia supervisor or your Sylvania supervisor?

Schreiber:

My Sylvania supervisor. This was before I went to Harvard. I really studied and absorbed what Shannon had in mind, and I thought I understood it quite well. I don’t think I fully appreciated the role of noise in this whole business, but nevertheless it came up in my thesis because of the character of the measurements that I had made. One of the people on my committee was Middleton, who was a theorist in a number of aspects of communications, and he caused me a lot of trouble. Because of him I had to re-write a part of the thesis where I was dealing with the effect of noise on the measurements that I was making.

Nebeker:

Who at Harvard was working on such matters?

Schreiber:

Well, no one really. No one that I know of. The philosophy of the place seemed to be to let the students do more or less whatever they wanted. I had a full fellowship at Harvard -- a Gordon McKay and a Charles Coffin Fellowship from GE, and so it didn’t cost me anything to go to Harvard. It certainly was a school with a very good reputation, not so much in engineering but in education generally and in science. I made the measurements, and I graduated. Some months before I finished up I was visited by a representative of the Technicolor Motion Picture Corporation from Hollywood who was scouting for people, and eventually I went to work for Technicolor.

Nebeker:

Could I go back just a minute to the Harvard experience? You said you were always interested in pictures. Is there more of an explanation there of how…?

Schreiber:

Well I worked steadily in photography from the time I was maybe 10 years old. I built my own enlarger. I didn’t know much about optics at the time, but I eventually learned.

Nebeker:

Your thought was that using the Shannon Theory you could make some contributions to the transmission of images efficiently.

Schreiber:

Yes.

Nebeker:

Were there people who had been doing that?

Schreiber:

No, not very many.

Nebeker:

And it was your own idea just because you were interested in images and you’d learned about this theory?

Schreiber:

Right. There was a fellow at Bell Laboratories who was doing some work along those lines. I don’t remember the details, but… No, it was mostly my own idea, and spending my professional life working on my own ideas seemed like a good thing to do.

Influence of Depression and World War II

Schreiber:

I was just talking with some people last night at dinner about this. I said that in that part of my life I was very optimistic, even during World War II. I never doubted that we would win the war, although it would not be easy. (My brother was shot down over Germany and was a prisoner of war for months before we heard that he was OK.) We had so much more industrial power than the Germans and the Italians. Of course we had a good supply of brave servicemen who did the actual fighting, but so did the Germans and the Japanese.

I felt that even though I was the first person in my family to go to college, everything was open to me. I could do anything that I wanted, and that made me feel really good. Last night we were talking about whether today’s kids have that feeling, or are to some extent worried about the future. I think at least some of them are worried and not as optimistic as I was at the time. Of course it was a very different world.

Nebeker:

I’ve talked to quite a few people who got their degrees right after World War II. Began their careers then. And I think that is a common feeling, exciting things were happening.

Schreiber:

You could do anything. And education was the key. Neither of my parents graduated from high school. My father eventually got an equivalency certificate, but my mother was basically uneducated. I think she left school in the eighth grade. But both of them believed that education was the key to success. As I said, in my mother’s view, success was to be a high school teacher. My father was a man of few words and didn’t express exactly what he thought success would be. I remember the day I got the notification about winning the Pulitzer Scholarship to go to Columbia. He kissed me for the first time in many years. In retrospect I have to say one of the things that made him so happy was he was not going to have to pay for my college education. My father had a great reluctance to spend money. (I suspect that living through the Depression may have had that effect.) For example, my mother always wanted a house and my father was against it. Well, later on I had enough money to buy them a house -- one of the best things I ever did. It made my mother so happy, and it turned out that my father was perfectly willing to have a house as long as he didn’t have to pay for it.

Nebeker:

Well, I think there are people who grew up in very hard times and just didn’t want to spend.

Schreiber:

Well, I was brought up in the Great Depression. I was born in 1925, so I became aware of the world around me during the depression; money was the main topic of discussion in my house. My father never lost his job but he was down to $25 a week. My mother’s table money was $12 or $13 a week. Anything expensive was served at lunch time where I could have it without my father seeing it. I was my mother’s favorite, although I now get along quite well with my older brother. Well, even then we got along quite well, as we were four years apart and there was not much direct competition. In retrospect, my brother was very tolerant of this. He and his friends let me tag along when they went skiing, for example. His social skills were much better than mine, and he was a good athlete and very popular with girls, so he had a life outside the family that must have given him a lot of satisfaction.

Nebeker:

Do you have other siblings?

Schreiber:

No, just the one brother. He retired as a Colonel in the Air Force. He’s a very different person from me, although we look alike to some extent. Anyway, where are we?

Technicolor employment; video bandwidth, film, and color science

Nebeker:

Then you were approached by Technicolor.

Schreiber:

While I was finishing up my PhD thesis, a sort of a scout came by from Technicolor actually looking for somebody working in the field in my field. The reason for that is that one of the members of the research department at Technicolor, a fellow by the name of Donald Kelly, thought that information theory was important to what he was doing. He died only last year. It was quite eerie. I was writing a paper for Proceedings of the IEEE about the Goldmark field-sequential color TV system. I called up Don to get some information, some dates and so forth because he knew about all those things, and there was his voice on the answering machine. A few weeks later I got a call from his son, whom I had met when he was a child, saying that his father had died. I was really shocked because Don was a friend of long standing and the person who introduced this whole field of work to Technicolor. He was the single person who had the most effect on my career.

I went to work there in 1953, and this was the period during which people started talking about video recording. Do you remember that Bing Crosby was involved in video recording? There seemed to be a big application in television and maybe in the movies. Don felt that a key to video recording would be bandwidth reduction because the video recorders at the time simply didn’t have the bandwidth required for video signals. As it turned out the video recording problem was solved, not by reducing the bandwidth of the signals, but by increasing the bandwidth of the tape.

Nebeker:

What was your dissertation title? Did it have some…?

Schreiber:

“Second Order Probability Distribution of TV Signals.”

Nebeker:

So by itself it wouldn’t have attracted the attention of industry.

Schreiber:

No, it wouldn’t. Anyway, this scout from Technicolor came and I went out there for a visit, and one thing led to another. One of the nice things was that they offered a very substantial salary. I had been dickering with Bell Laboratories, which obviously was a place definitely suitable for me, and they offered me $6800 or something like that. I complained. I said that was not enough, because I had offers from other companies for quite a bit more. So they raised it, and they claimed it was not in response to my letter, but they were raising all their offers. In the movie industry at that time everybody had employment contracts. A friend of mine was a lawyer and I had him negotiate the contract. I started at $10,000 and I got a 15% raise every year for five years.

Nebeker:

It was a five year contract?

Schreiber:

A five year contract. I ended up working there for six years. We moved to California and I started working at Technicolor. At that time our first child had come along.

Nebeker:

Can you tell me a little bit about what was going on at Technicolor?

Schreiber:

Audio File

MP3 Audio
(345 - schreiber - clip 1.mp3)

Technicolor was in the business of making color motion pictures. They had the first practical process for making three-color motion pictures, using the dye-transfer process. The process was quite old, but they were the first ones who succeeded in making it work for small-format film, which was no mean engineering task because you have to do things at very high speed. The film went through the transfer machine at 2,000 feet a minute or something like that. You couldn’t see it, it went through so fast. The accuracy of the alignment of the three images on film -- about 1 or 2 ten-thousandths of an inch -- was the kind of thing that nobody has ever achieved before. Kodak provided the film, and it was Kodak’s opinion that it would never work because they didn’t make the perforations accurately enough to do this kind of registration. Nevertheless, Technicolor made it work. Looking back on it, it was a rather crude open loop system --no feedback or anything like that. The way they got their registration accurate enough was pin registration. There were three pieces of matrix film that held the yellow, magenta, and cyan dyes to be transferred to the receiving sheet. They sandwiched the matrices one after the other with the receiving sheet on a pin belt. The pin belt was a 35 millimeter wide strip of monel metal with pins that fit the perforations on the film. The film was impaled on these perforations over a distance of several feet, averaging out the errors in the positions of the perforations. Dye transfer occurred when the receiving sheet and the matrix that had carried the dye were sandwiched together on the pin belt, and passed through a hot chamber.

My ostensible job was to apply information theory to video bandwidth reduction. For this purpose, we designed and built a “probability machine” and measured some more probability distributions. I also started working on some aspects of the film business. I learned a lot about color. Don Kelly at that time was doing his Ph.D. thesis at UCLA, and his field was color perception. He made very significant contribution in that field. Later on he worked at Stanford Research Institute doing the same thing until he retired. He is well known in the psychophysics field for his work in that area, but he was also more interested in the more practical things. Anyway, I learned about color.

Nebeker:

But all of this was film-based technology. Is that right?

Schreiber:

Yes. Interestingly enough, at that time Technicolor Corporation had made a contract with Stanford Research Institute to build a simulator, a television-based simulator that they could use printing the color film. There are three variables in the printing process they were using -- the three exposures of the matrix films to the original negatives. In the Technicolor camera they didn’t use color film, they used three strips of black and white film that went through camera. The image formed by the lens was separated into three images using multi-layer dichroic beam splitters that formed the red, green, and blue component images. If you visit the Eastman Museum in Rochester, you can see one of original Technicolor cameras. It was a triumph of engineering.

The beam-splitting filters, which were made up of many thin layers of two transparent mirrors with different indices of refraction, were deposited on the surfaces of four triangular prisms, each with an apex angle of 90 degrees.…I know this won’t get recorded on your tape [drawing a diagram], but here is the lens and here are the prisms. The whole assembly was made so accurately that when you put the four pieces together to form a cube, you couldn’t see the line in the middle. The net result was a glass cube, called the “x-prism,” with imbedded crossed beam splitters. White light came from the lens into one surface of the cube and the three separate color images were formed on the other three surfaces of the cube. Three strips of film ran over those three surfaces, thus exposing the color separations right in the camera.

The Technicolor process made spectacularly good color pictures, and it is still used. They don’t use the camera any more, but the pictures are still printed by Technicolor in Britain using the dye-transfer process.

One of the people in the lab was designing and making these multi-layer interference filters using a punch-card computer. Some of these had as many as 39 layers. There was both art and science in designing and making the filters. Of course, it is much easier to make these today. In some cases they’re even wedged from one end to the other to take account of the fact that the light doesn’t go through the filter in parallel rays, it diverges. In any event, the separation negatives are exposed in the camera and developed, and from the resulting negatives they make the three matrices that absorb the dyes. The variables in the process are the three exposures of the matrices. SRI built a simulator that had three knobs on it, and these three knobs represented the exposures of the three layers.

Nebeker:

What kind of simulator? Analog electronics?

Schreiber:

It was an analog television system using analog circuits that simulated every aspect of the process. The way the simulator and processor were standardized was by the use of a set of standard separations that were processed in the system every day. The result was displayed on a monitor next to an optical projector that displayed a perfect print. The pictures were about 6 x 8 inches. The monitor was made from three cathode-ray tubes that were viewed in superposition through a set of beam-splitters. In a perfectly dark room from a few feet away you couldn’t tell which was the optical projector and which was the television. It was an excellent system.

Nebeker:

And this was in order to better design such beam-splitters?

Schreiber:

No. The purpose was to get the exposure right so you could produce pictures that had the right color. It turned out this is difficult to do, especially on scene changes where your eye can detect very small errors in color balance. Using this system, they would make a set of test exposures, and then display the result. By turning the knobs until a match was achieved, a corrected exposure was found. Generally, the second try would be acceptable. It was a tour de force; it made it possible to produce high quality results at a reasonable price. I contributed very little to this system, as it was mostly developed before I came to Technicolor. I did work on other aspects of their color printing system, which was quite educational for me.

For me, the most important element of my stay in California was that Don Kelly tutored me in color science. By the time I left Technicolor I was an expert in color and that came in very handy later on.

Comparison of industrial and academic research

Schreiber:

Peter Elias, who had been on my thesis committee at Harvard, was a Junior Fellow in a the Society of Fellows at Harvard. It is a great honor to have such a position. The Fellows are recent graduates but it is not a path to joining the Harvard faculty, as I learned to my disappointment. When I was close to graduation I went to the Dean (Van Vleck) at Harvard and said, “Gee I’d like to teach here.” He laughed and laughed. He said that is not the way we hire faculty. He said when we have a faculty opening we appoint a committee and they look at the whole world and select the best person. If you want to teach at Harvard, go somewhere else and make yourself a record.

While I was working at Technicolor I kept my connections back east, making a trip at least once a year. I kept up with Peter Elias, who had become a faculty member at MIT. He urged me to come to MIT, which had never occurred to me. It seemed like a trade school compared to Harvard. But after six years in Hollywood I moved to MIT as an associate professor without tenure. I uprooted my family again. By this time we had three children.

Nebeker:

And you made the move because it seemed a lot more interesting to be doing research rather than what you were doing?

Schreiber:

Well, yes. There were many strictures about working at an industrial place like Technicolor. The president and founder of Technicolor, Herbert Kalmus, had very old-fashioned ideas. (Technicolor was named after MIT. It was originally located in Cambridge.) For example, I wanted to teach in the evening at UCLA; I was refused permission on the grounds that I was working at Technicolor and I should devote my life to Technicolor. The next term I just started teaching without asking permission. I looked forward to coming to MIT because I could pretty much do whatever I wanted and I’d have an opportunity to do consulting and other things. I took a big drop in salary. At that time, with my employment contract I was making $18,000. I went to MIT for $8,300 for nine months. I started consulting, but it was not easy to get assignments at first. I was getting $50 a day, which was considered not bad in those days. I charge a lot more than that per hour now. MIT on the whole was a very good experience for me, even though the salary has never been comparable to industrial salaries.

MIT teaching and research; imaging systems

Nebeker:

What research did you… Was it something you had been working on at Technicolor that you…?

Schreiber:

When I came to MIT, I was required to teach my way through the core curriculum, and that’s a really good education, but it takes a lot of time and effort. Columbia was a back water in education at the time I went there. The curriculum hadn’t been modernized for ages. Of course it was during the war. MIT, on the other hand, was constantly modifying its curriculum. They had gone to a science-based curriculum. I found the first few years teaching my way through the core to be very, very difficult. I had to put an enormous amount of time into learning that stuff, but it was good for me.

As for research, I didn’t have any fixed idea about what I ought to be working on. Somebody told me this was the day of the Esaki diode: why didn’t I become an expert in negative resistance? I actually looked at that for a few months, and it just seemed like the most boring thing in the world. Teaching the core kept me busy for the first few years. Finally, I got to the point where I could come up with my own course, and I decided that I wanted to teach image processing. I made a proposal that was accepted, and I started teaching a course called image transmission systems. Eventually that grew into a graduate course that I taught until I retired, and from which I produced my only book. The course was also useful in recruiting thesis students.

Nebeker:

Was this all types of imaging systems?

Schreiber:

Yes. One of the things I concluded at an early date is that all imaging systems have a lot of things in common. In recent years, I have often served as an expert witness in patent litigation. That concept has been very useful to me in court because the lawyers often say, “Well this isn’t about television, this is about facsimile.” Then I go into my speech about common elements.

Nebeker:

But your course was about all kinds of optical systems and electronic systems as well?

Schreiber:

Audio File

MP3 Audio
(345 - schreiber - clip 2.mp3)

Yes. Well, I spent a lot of time in the library getting this course ready. I went to the New York Public Library and the Boston Public Library and found all sorts of interesting old stuff. In those days everything was on the shelf, so it was easy to find. For example, I found that in the ’20s there was an amateur television society in Boston where people were building their own television sets. Some companies were making kits. I just found it tremendously interesting. I mentioned some of that historical stuff when I started to speak today because most people really don’t know about it. Who knows today that facsimile is older than the telephone? But it is.

I was a faculty member at MIT from 1959 until 1990 when I retired. I got tenure in 1968, and eventually was promoted to full professor. For a few years I was the Bernard Gordon professor, an honorific post where I was trying to get some more practicality into the curriculum, which had become exclusively science-based. I have very strong views about engineering education that are at variance with the views of most other faculty, so I’m very happy that I don’t have to worry about it any more since I’m no longer a faculty member. I can go to faculty meetings and I can talk all I want, but I have no responsibility in that area.

When the science-based curriculum was first introduced at MIT, it was taught by engineering professionals. These were people who had an immense amount of practical experience building things, so they could do a very good job. However, when a science-based curriculum is taught by people who graduated from a science-based curriculum and have no other experience, the contact with reality disappears. I think that has had very serious consequences for the country. Places like MIT attract the very best students, but the level of engineering quality in the United States is not very high. That is why bridges fall down and why American cars aren’t as good as Japanese cars, although ours are getting better. This is because we don’t have the best people doing that. The best people do the best work. But in the United States at the present time, the best people go into other fields. They don’t go into engineering. They become lawyers. Of all the professions I deal with, (I’ve spent a lot of time with lawyers) the average intelligence of the lawyers is higher than that of any other profession. Now that’s too bad because a lot of the work they do is nonproductive. And then a lot of very bright people go into medicine where you can make an immense amount of money, and a lot of bright people go into business. They get MBAs and they pull down a lot of money. I know one young kid who went to law school, and not a top flight law school. He started off at a good firm at $70,000. He is now well over $100,000, with prospects for doing much better in a few years. Quite a few of my own former students are making a lot of money too, but they have much more education than the lawyers have. They’ve all had eight, nine years of school and image processing is a very hot field. Students who have specialized in image processing and come from good schools can get very respectable salaries. However, the people who design the mundane products, the automobiles, the bridges, the roads and so forth, they’re the Wentworth Institute graduates.

There is nothing wrong with kids who go to schools like Wentworth, but they are not the best students. It’s just too bad. I don’t see that situation changing at all. My efforts to introduce some reality into the MIT curriculum were a total failure. I really put my heart and soul into it for three years just because I though it was important, but my colleagues totally rejected it.

Practical research applications and industrial sponsorship

Nebeker:

I’ve heard people in signal processing comment that your work seems always tied in some way or at least consistant with practical systems.

Schreiber:

My laboratory, unlike most at MIT, was industrially sponsored almost the whole time I had it. I learned how to work with industrial sponsors. There is some difficulty in balancing the needs of the sponsor with traditional academic concerns such as complete freedom to publish. After a while, I worked out a modus vivendi. For example, when we were developing the facsimile system that the Associated Press now uses (they call it Laserphoto), the United Press never heard a word about what we were doing until we made a public announcement. It wasn’t very difficult. I didn’t have to issue any instructions to the students or anything like that. They knew what I had in mind.

Nebeker:

Are you saying it was intentional to not publicize this work?

Schreiber:

Yes. It was intentional.

Nebeker:

Was that required by AP sponsorship?

Schreiber:

Not exactly. But you know, it’s like when Ford pardoned Nixon. When Ford was appointed to take Agnew’s place, I’m sure he and Nixon didn’t have any discussion about a pardon, but they both came away from the meeting knowing that that’s what would happen. Did you read the op-ed by Ford in the Times yesterday? It’s kind of interesting because he refers to that, but not in terms I just mentioned.

In order to understand this situation, you have to look at the reasons why companies sponsor work at universities. Some companies do it for prestige. There are AT&T fellowships for example. They don’t really care what the students do. In the case of a company like Providence Gravure, a company that spent $2 million in my lab at MIT, they expect the work to do something positive for their company. The same thing applied to the Associated Press. The students readily accept this reality. They find it exhilarating to be competing with professionals. I think it is fully possible to balance the students’ needs with the sponsors’ needs without destroying traditional academic freedoms, as long as you don’t try to write down the rules.

Because of the fact that you couldn’t go out and buy proper equipment to do image processing at the time I started doing serious work in the field, I became interested in the apparatus. E.g., I became quite knowledgeable about scanner design. We designed and built a CRT scanner where the sensitivity point by point over the screen was uniform to well within one percent, using feedback. After we had done the most that could be done with CRTs, I got interested in laser scanners, and we were able to make substantial improvements while keeping the costs down.

Image processing research; noise visibility

Nebeker:

Can you tell me sort of in order these projects that you undertook at MIT? You said initially you were concerned with core curriculum. How did you get into real research on image processing?

Schreiber:

I was in the Research Laboratories of Electronics (RLE). In gratitude for the development of radar during World War II, RLE was supported by the Joint Services to the tune of several million dollars a year, with complete freedom to do whatever we wanted. A most remarkable situation. I joined an existing group led by one of my colleagues, Sam Mason, who unfortunately died early from a cerebral hemorrhage. Sam was interested in sensory aides for the blind and had started a project to build a reading machine. You put in a book and it would speak the words. They needed scanners for that, so I got the scanner concession. I ultimately became an expert in scanners. I built all kinds of mechanical scanners, CRT scanners, and finally laser scanners.

Nebeker:

Did you patent that work?

Schreiber:

I wasn’t very patent conscious in those days. For example, our feedback technique for getting high uniformity across a scanner, which is now called calibration in the printing field, definitely would have been patentable and probably would have been worth something, but I didn’t think along those lines in those days.

Nebeker:

So you published your work publicly?

Schreiber:

We published everything of interest; we didn’t try to hold anything back.

Tom Huang was one of my early students, and he worked on noise visibility, which opened up an entire new field.

Nebeker:

Noise visibility in a particular system?

Schreiber:

In pictures. He investigated what the spectrum of the noise in the picture had to do with its visibility. That was a basic scientific approach to the problem.

Synthetic highs, compression, multi-channel coding

Schreiber:

Then he got interested in compression. You recall that I had been working on compression at Technicolor, where we developed the “synthetic highs” system, which a number of other researchers referred to--

Nebeker:

I’m sorry. I wonder if we could go back and talk about that because that paper, I know has been influential.

Schreiber:

Well, it’s been known for hundreds of years that a picture can be represented, at least in part, by outlines. In fact if you go back to some of the cave drawings in France, which are 25,000 years old, you find they’re outline drawings. When kids make pictures in sand with a stick or with their finger, they’re making outline drawing. This is a representation, a kind of representation. I began to think about this, and I devised a scheme where if you knew not only the location of the outlines but something about the brightness change that occurred at points along the outline, you could reconstruct a good picture. To get good compression, you would have to represent the outline in a coded fashion. At that time, I don’t think I was fully aware, of the work that had been done in typography using outline coding. Tom and I and one of our other students, Oleh Tretiak, wrote a paper on contour coding.

Nebeker:

When was this, roughly?

Schreiber:

It was in the early 60s. I thought it was a very good paper. It has been included in several collections of significant papers.

Nebeker:

Before you went to MIT?

Schreiber:

No, it was while at MIT. The exact dates are in my CV. That paper was reprinted in a number of collections of papers because we pointed out that it was a coding technique that was in-between certain other techniques that had been proposed at the time and had the possibility of very high compression if you could figure out how to do the code for the outline economically. The system separated the spatial low frequencies from the spatial high frequencies. What we discovered was that the best thing to do was not to try to reconstruct the entire picture from the outline, but only the spatial-high frequency component. The spatial low-frequency component could be transmitted without coding because it didn’t take up much bandwidth. So that was the beginning of multi-channel coding, although there had been a paper by E. R. Kretzmer from Bell Laboratories, showing the benefit of dividing a video signal into spectral components and treating different spectral components differently.

Nebeker:

But the Synthetic Highs paper, that was before you went back?

Schreiber:

Yes, that was at Technicolor. That’s where I became interested in the low frequency high frequency business. Eventually, I met Kretzmer and we were always on very good terms. Of course, I had given him proper credit in my published papers -- I have always been very careful about that kind of thing.

The synthetic highs system didn’t make perfect pictures. One of the reasons was that not all of the information in pictures can be well described by outlines. To achieve any worthwhile compression, you have to have a threshold for what you call an outline. If there is a small change below this threshold, you don’t call it an outline. The threshold determines the compression. If you have a high threshold you get high compression; if you have a very low threshold you don’t get much. In the first case, the reconstructed picture has defects, while in the second case, the pictures are much better.

Nebeker:

It’s kind of like these paint by number systems. Sometimes you see zillions of little words.

Schreiber:

If you get the threshold so low that you don’t miss anything in the picture, you don’t get any compression. Murat Kunt in Switzerland tried to solve this problem later by coding a third component: lows, highs, and random components. I visited there at the time and they were not getting very good pictures. But anyway, that became a side issue for me. Once I got to MIT, I didn’t really work on synthetic highs any more. I worked on more practical problems.

Nebeker:

What was the hope when you wrote that paper?

Schreiber:

Well, the hope was that we could very high compression ratio.

Nebeker:

Was that something Technicolor pursued?

Schreiber:

Oh, sure because they were interested in video recording, and they thought compression was one way to do video recording, an idea that was ahead of its time. Compression really hasn’t been used in commercial systems until very recently. One of the Japanese companies made a 45 megabit per second system for NTSC that was commercially used to some extent. It’s not perfect, but it’s quite good.

Facsimile transmission for the Associated Press

Nebeker:

So, if we could return to the early ’60s probably at MIT when you’re beginning your building up the program there.

Schreiber:

We were contacted by the Associated Press to do a job for them. They had gone to Bell Laboratories and had run into one of my former students, and he had recommended that they come to MIT, since Bell Laboratories couldn’t do what they wanted, or wouldn’t do what they wanted. Their problem was facsimile. The AP runs a world-wide facsimile network. The facsimile transmission standards then used in the US and the rest of the world were different. The procedure used by the AP was that when they received foreign pictures in New York they made paper copies, put them on an American machine and rescanned them and vice versa when they were transmitting from the United States to Europe. You don’t get very good quality that way. They wanted a high quality scan converter that would be located in New York City, which is the hub of the AP universe. At that time RLE was not very amenable to doing such practical projects, so my colleague Don Troxel and I did it as a consulting project. One of my former students had an electronics company near MIT, so we did the construction there.

Nebeker:

Approximately what year was this that you started the AP project?

Schreiber:

We started that project in 1971 or 1972, and it was finished in 1974. One thing the Associated Press knows about is publicity. That’s their business. When the system was announced, it got immense publicity, including a long article in The New York Times with my picture looking through optical system. It was a very good system.

Electronic Character Recognition Machine

Optical character-recognition (OCR) machine

Schreiber:

At that time I had my own company, ECRM, which was founded to do optical character recognition. The company was formed as an indirect result of the consulting project that we had done for the AP. One of the people at the AP who was involved in that desperately wanted his own company. He came to us and said we should start a company and he would be involved with us. His goal in life was to make $100,000; he was from Texas and he wanted to have a small cattle ranch, and he could do it with 100K. He did not really care what we did as long as it was likely to be profitable. We decided to develop an optical character-recognition (OCR) machine to be used in the printing industry. I mentioned the reading machine project earlier. It had gotten to a point where--

Nebeker:

And you were working on the scanner for the reading machine project?

Schreiber:

Yes. The reading machine project had gotten to a point where the character recognition was coming along pretty well. The speech generation was not, but the character recognition seemed to us to be useful by itself, so we decided we would build an OCR machine for the newspaper industry. That was the Genesis of our company.

The facsimile scan converter that we had built earlier was very successful. The machine that we built and installed in New York was used for decades. It has subsequently been replaced, but the basic system is still there. There is a company in Framingham, Massachusetts, called Leaf Systems, which inherited the project. The technology in the newspaper business has now gotten to the point where everybody is using computer systems to do pictures. Pictures are now transmitted by the AP directly into these computer systems. They don’t really need the facsimile machine any more, but the basic idea is still there. We never made any money out of it except for our consulting fees, which in those days were quite modest. Now I charge what the traffic will bear because I’m not looking for extra work.

Nebeker:

So how did new company do?

Schreiber:

It was called ECRM, which meant Electronic Character Recognition Machine. Again I did the scanner and my two MIT colleagues worked on the OCR. We went through a series of scanners -- first with vidicons. We had six camera tubes placed across the page so they could do a very high resolution scan of the page, and then we had a computer system that did the OCR. First a PDP 8 and later on a PDP 11. Ancient machines, but they worked. At one time we were the largest user of the PDP 8.

Nebeker:

Did you sell systems?

Schreiber:

Oh yes. The company was quite successful for a while. We eventually fell on bad times and we sold the company to Addressograph-Multigraph. I had earlier started another company specifically to do facsimile data compression, and that company had earlier also been sold to Addressograph-Multigraph. I actually succeeded in selling two little companies to AM. In each one I made what seemed then quite a lot of money. It was kind of nice.

Nebeker:

Who was interested in the OCR machines at that time?

Schreiber:

Well, our plan was sell it to newspapers. That was our market, first for classified ads and then later on to other things. We were competing with video display terminals, which were just coming into use at the time. With the OCR machine, the operator could take the ad over the telephone on a Selectric typewriter. We provided for some hand editing that the OCR machine would recognize, such as correcting typos and crossing out letters and words. We placed quite a few of those machines. The unions didn’t cause us any trouble mainly because they didn’t think the system would work. By the time it was in and working it was too late. There was no way they could reverse what we had done. (In retrospect, I am not too proud of this kind of de-skilling.) Later on I began to have second thoughts about displacing workers, but at that time in my life -- let’s call it a youthful indiscretion -- I didn’t have as much social concern as I have now. The second company, ECRM, is still in existence.

Autokon word processor

Schreiber:

After we made the OCR machine, some of my colleagues wanted to make a word processor, and I thought that was about the world’s worst idea. I had done a count and discovered there were 27 companies making word-processing systems at that time. I said that even if we produced a better system, we would never make a dime. There is no way you can compete against big companies making word processors. I suggested instead that we make a scanner to make halftone pictures to be used in newspapers -- something that could enlarge and reduce and adjust contrast and brightness and make a halftone picture ready to be pasted up in newspapers. No one else had a fleshed-out proposal, so my suggestion was adopted. We called the product the Autokon. It was fabulously successful in that it introduced an entirely new device that was of real value to its users.

Nebeker:

You set up a company?

Schreiber:

No, we did it at ECRM. It didn’t require a new company. We had enough money from the OCR work to develop the system. We eventually produced something that worked very well, and it was relatively inexpensive. This was a project that gave me immense satisfaction. I was able to use absolutely everything I had ever learned about image processing, and also about electronic and optical design and construction. I had now become a pretty well qualified optical engineer from the various other projects that I had worked on. We got a patent from which we never made much money, but the machine itself, which we called an electronic process camera, eventually became widely copied. No one built a machine anywhere near as cheap as the one we had. So it was--

Nebeker:

When was this done?

Schreiber:

It was done in the mid-70s. The AP project was finished in ’74, and I had gotten a lot of experience with laser scanners on the AP project. First I applied this experience to make a laser scanner for our OCR machine, and then a laser scanner and laser recorder for the halftone machine. Those are projects that I really enjoyed because I was able to use every thing I had learned from my earlier work on scanners. I cheerfully devoted an enormous amount of time and effort to that project.

Color printing for Providence Gravure Company

Schreiber:

As a result of the publicity that surrounded the AP Laserphoto project and the Autokon development, we were approached by the Providence Gravure Company to develop a complete pre-press system for color printing. This was long before the development of page composition systems such as are commonplace today, such as the Aldus Page Maker and the various systems from Adobe. Just as importantly, there were no computer-based color printing systems, although there had been some early proposals.

Providence was printing a lot of advertising, but they also did the magazine section for the Boston Globe, the Washington Post, and theProvidence Journal, which actually owned Providence Gravure. The process used for this color work, although it produced very good results, was slow and expensive, involving many makeovers and a high degree of skill on the part of experienced craftsmen. They dearly wanted a computer-based color printing system that would save them time and money.

Nebeker:

This is to replace all the old cut and paste kind of composition?

Schreiber:

Audio File

MP3 Audio
(345 - schreiber - clip 3.mp3)

Right. In the old system the original pages consisted mostly of text and photographs that were cut and pasted. The pages were then put on a big scanning machine and scanned. The analogue signal from the scanner went to an engraving machine which actually made the gravure cylinder. The main idea of the system that they wanted was to make the gravure cylinder directly from the computer. The components for the pages would be individually scanned into the computer and the page layout would be done there.

My colleague Troxel did the page layout system and I did the color system. By this time I really understood color. We produced a spectacularly good system that was installed in Providence. We made sample pages that were printed in the Providence Journal. The color was excellent, and moreover we got good color the first time, without makeovers. The key to that was we had a CRT simulator that accurately predicted what the page was going to look like.

By that time I had become quite patent conscious. Remember that this was an MIT project not a consulting project. We never got a penny of consulting pay out of it as we had on the AP facsimile scan converter. I did think that there was a possibility of making money from the novelty and usefulness of the Providence color system, but the MIT patent department didn’t want to pay for doing a patent. They had done quite a few patents for me without much return. The AP patent had made some money, but that was only by accident because the AP thought they could resell the system to other people, so they started paying for their own use.

I was convinced that our color system was much better than anything then being used because it was cheaper and faster. I had now studied very carefully how other people were doing it, and this was much better. It didn’t require highly skilled people to operate. Good color work could be done by a computer operator with only a little training. The operators did have to look at the picture on the screen carefully. We gave them knobs with which they could easily change practically anything on the picture, and immediately see the result on the screen. They could change the individual colors one by one. They could change their hue or their saturation or their brightness. The traditional color people at Providence loved the color system but they didn’t like the page layout system because on average it took them two and a half hours at a terminal to lay out a page.

In the mean time, I had visited Japan where a large newspaper, the Asahi Shimbun, had built such a system. It required an entire new building and was extremely expensive.

Nebeker:

And this was independently?

Schreiber:

Yes. There was quite a lot of work going on in Japan along similar lines, but their costs were absolutely astronomical. With my penchant for making things cheap, in which I now had a lot of experience, I thought they were doing it all wrong. In the opinion of the Providence management, our system pretty much did it all right, except they thought two and a half hours was too long to lay out a page. As a result, Providence gave up the system. They had spent about $2 million at Providence at that point and $2 million at MIT. $4 million is quite a lot of money. Because of their dissatisfaction with the speed of the layout system, they made a deal with one of the big scanner producers in England to jointly build a new plant based on the English system.

Although I was very disappointed that Providence decided not to use the system that we had developed, I eventually got some satisfaction out of this. I remained on good terms with one of the executives at Providence, and when they finally got the English system in and working, I said, “Paul, how long is it taking you to lay out a page?” He said, “Two and a half hours.” Nobody did the color as well as we did. The MIT patent department didn’t want to do it so I said I’ll write the patent specification, but I need a lawyer to help me with the claims. This patent has now been licensed by almost every major company in the color business.

Color printing patent and royalties; EFI

Nebeker:

Is it you or MIT who's the patent holder?

Schreiber:

Audio File

MP3 Audio
(345 - schreiber - clip 4.mp3)

MIT is the patent owner, but we all made money out of it. MIT has a very generous patent policy. The inventor gets a fraction of the royalties. Providence also got a fraction of the royalties because they had some interest in the patent due to their sponsorship of the project. The main reason why the patent has been so successful is that one of my early students, an Israeli named Efi Arazi, had founded in Israel a company called Scitex to do computer based color work, initially for the textile industry and later for newspapers and magazines. Their approach was much less radical than ours. They were essentially using all the old expertise of the scanner operators. They did not do color correction in their system. They did a very good job on page layout in the computer and they had all sorts of tricks about taking the page components and putting them together properly. It was a great success, although quite expensive (and correspondingly profitable) because it was built around a special-built computer system that cost about a million dollars a copy. With the development of PCs and Macs, Scitex was put at a distinct disadvantage. Especially when the Macintosh came out, the handwriting was on the wall. They could not do this any longer. They were forced to run their system on cheap personal computers and almost went belly up.

Efi Arazi left Scitex and moved to California. He married an American plastic surgeon and decided to start a company in San Francisco based on my color patent. I helped him make up his story to the venture capitalists. He is not only a natural-born engineer, he is the world’s best salesman; I think he could sell ice to Eskimos. He raised enough money to get started and gave me a small piece of the company. I also served as a consultant.

Efi named the new company after himself, Electronics For Imaging. EFI became a spectacular success, but not based exclusively on my color work. EFI developed a software version of the Providence system for desktop publishing that got all kinds of prizes. The most innovative system, etc., etc., which it was. I had originally intended to revolutionize high end printing. Now high end printing has in fact been revolutionized, and pretty much along the lines of my work, and some royalties were collected. However, it was in desktop publishing that most of the royalties were collected. Efi took an exclusive license to the MIT patent and has now collected more than $16 million in royalties. The first big check was $1.5 million from Kodak. I had a copy on my wall for a long time because I had tried to sell that patent to Kodak on behalf of MIT. If they had given me $25,000 I would have been delighted, but they ended up paying $1.5 million.

Efi was very clever about his approach to the licensees. He was not greedy. He did not ask for large amounts of money. Now that I do expert witness work I realize if you go to a company and you ask for a million dollars, they will almost always give it to you, no matter how good or bad the patent is, because that’s cheaper than going to court. If you ask for $10 million it’s something else. If you ask for $100 million no one will pay without going to court. Efi knew exactly how to do this, and he signed up every large company in the color field. The last one to sign was Apple. MIT gets a certain fraction of the royalties, and I get a fraction of MIT’s fraction. I also had stock in EFI, so it worked out quite well for me. It made me very happy that this was all legitimate money earned out of my own head on a really good technical development.

Nebeker:

And this technical development grew on all the more theoretical work.

Schreiber:

Yes. The success was based on the color science that I had learned at Technicolor, my analysis of the processes needed to make good color prints, and also on my views on how to build complicated machines to make them reliable and at the same time relatively inexpensive. I also considered very carefully which operations should be done by the computer and which by the operator.

Nebeker:

Now it sounds like a lot of your work has been on the device side. Scanners and etc. Is that right?

Schreiber:

Oh yes, I spent a lot of time on that. But with respect to color work, the first part of the job was planning the theory of operation -- how the system should process the information without regard to the hardware and software that would be used to implement the process. There were a lot of approaches that different companies were using. I rejected all of them on the basis that they were too complicated and hard to learn. They required too much expertise on the part of the operator or they wouldn’t give very good color or the picture quality -- the resolution and sharpness and the freedom from defects -- wouldn’t be what the people in the field would accept in view of their very high standards of quality. I skipped over an important piece of the success story -- a part that was eventually very important to me. After a while, EFI discovered that they had got the economics of desktop publishing all wrong. They should have known this in advance, but they failed to talk to people who knew that field well enough. They were considering hiring a marketing person at one point, and I had a long talk with her. She said, “You know, in this field if you sell 20,000 copies of a program you’re doing really well.” There were only one or two companies that were making good money in desktop publishing at that time. One of them was Adobe, which had developed a rather good system that allows you to do the whole pre-press job, while we were focused on just a fraction of the job. The EFI system did not do page layout, just the color processing. Even though the color system was better than anyone else’s, it would have had to be incorporated into something like the Adobe system anyway to make it useful. The volume would be not nearly high enough to support a company that would fit Efi’s idea. He wanted to make tens of millions of dollars, and there is no way you could make tens of millions of dollars in desktop publishing.

Something else happened that made EFI extremely successful. Efi had hired a bunch of Israeli engineers who reverse-engineered the Canon color printer. That printer, unlike most Xerox machines, most copying machines, has a completely separate scanner and recorder that are joined by a cable that carries video signals. You chop the cable, you put in a computer, and now you have a system that has a lot of possibilities. What EFI did was to design and start selling controllers for color printers of this type. They now have a good part of the world market for printer controllers. Just a second. Are we finished here?

Nebeker:

Probably close to it. Let me get this.

Schreiber:

They developed a controller for color printers, and color printers have now become ubiquitous. They aimed initially at very expensive color printers, like the Canon color copier, but I think they now have systems for rather inexpensive ones as well. For a while they had pretty much cornered the world market. I calculated at one point that if their success continued they were limited to a total sales of about $5 billion a year, and many people would not regard that as much of a limitation.

I told you earlier about my business associate who just wanted a $100,000 ranch and quit when he had reached his objective. Well, my goals were a little bit higher than that, but I did meet them as a result of my work in color and in scanners. I can also take satisfaction in the fact that my work has had a substantial effect on the practice of color printing. Although I am still doing some work in the field, I’m beginning to lose interest in that entire area. I was at it for many years and I’m now much more interested in economics and social policy. That’s what I spend my time on. I really want to be a columnist, but I don’t have anyone yet who would publish my columns on a regular basis. I have had quite a few letters to the editor published in good newspapers, but not as many as I would have liked.

Image processing as a discipline

Nebeker:

There was a comment you made yesterday that I wanted to ask you about. You said that image processing wasn’t a science but an application.

Schreiber:

Right. I still believe that.

Nebeker:

Would you expand on that?

Schreiber:

Well, it’s not a branch of mathematics. I don’t think it’s even a branch of computer science. I don’t think there is such a thing as computer science and I don’t think there is such a thing as artificial intelligence. They are also applications. Image processing has become important because many users want their images processed and because that processing is done today on computers.

Nebeker:

Right, but when there is a large body of techniques, doesn’t that count as a science or a discipline?

Schreiber:

It’s a discipline certainly. A technology certainly. To science I give a somewhat more glorified definition.

Nebeker:

A more coherent theory.

Schreiber:

In a true science, as distinguished from a field of technology or a collection of methods, there has to be a relatively small number of principles that you can talk about, and I don’t think there is a small number of principles that you can talk about in image processing. That doesn’t make it less respectable. This is a reversal of the views I held when I was much younger. When I went back to school after working at Sylvania I wanted to be a scientist. I said I’ll be a physicist. I was wrong. That is not my forte. I’m an engineer. I like to design processes and products that work and do something useful, and that incorporate some science in them, thereby being better products. They do things better or they’re cheaper and more convenient, or more socially useful. I think that is highly respectable. One of the things I hold against MIT and similar places is that for many years developing practical products and processes wasn’t considered respectable. As I put it on a couple of occasions, if you’re in any high-class technical university in the United States -- when I came to MIT it was the only one, but now there are a dozen that are just as good and almost as prestigious -- if you spend your time teaching your students how to build things, pretty soon you will have to look for a job elsewhere. If you want respect and tenure at today’s technical universities, what you want to do is write papers, and the more theoretical the better. This approach, of course, has been taken in image processing, and most of the work so produced is totally worthless.

Academic vs. industrial engineering

Nebeker:

That is a point I wanted to ask you about, whether there is this danger that so much academic work is so divorced from the applications today that it isn’t very valuable?

Schreiber:

I think that is a real problem. As I say, that is one reason why the quality of engineering in the United States is not as high as it ought to be.

Nebeker:

Is that because the most talented people are being attracted to academia where they do these more theoretical things?

Schreiber:

Yes. We still get the very brightest students at MIT. For a long time the students who came to MIT were highly motivated toward engineering. They had been building things in high school, or as hobbies. That’s less true now because the bright kids spend all their time in front of a keyboard. I think that is a serious danger. I remember when the American automobile industry was getting into so much trouble with the Japanese. You’re not quite as old as I am, but you must know that Japanese cars were perceived by American buyers as more reliable, which they were. The reason is that the people who design cars in Japan are the best graduates of the best schools. Sony was one of my sponsors for a while, so I got to know a lot of Sony people. I once went through a Sony plant, and it was very educational for me. At that time, Sony was hiring new grads from Tokyo University, which is in a class by itself, or any of the second-rank schools that are also very good. Those students have nowhere near as good a graduate education as they would get in an American school, but they are nevertheless the top students. Sony puts them on the assembly line. They spend a couple of years where they really learn how things are made before they start designing anything on their own. That is an absolutely invaluable experience. We don’t do that.

Nebeker:

It seems to me that there are at least two problems here. That the brightest people get attracted to things like law and medicine and so on and maybe science. And the ones who are in engineering get attracted to academic positions rather than ones where they’re building the systems.

Schreiber:

Yes. Learning how to build things that work is something that’s not done by our very best students, and that is why bridges fall down.

Nebeker:

But isn’t there-- I mean, an outsider might think these people in academia working on image processing might be building up that theory that would make it a coherent science.

Schreiber:

Well, possibly. I hate to predict the future because I’m often wrong, as everyone else is. But it doesn’t look like that to me.

Nebeker:

You think it’s more important to be in touch with the present system and service.

Schreiber:

Absolutely, absolutely. I had a funny experience once. I had 33 or 34 Ph.D. students and maybe 75 or 80 master students and a couple of hundred bachelor thesis students in my time at MIT. I can illustrate this with a story about two of my thesis students. I can’t remember if they were bachelor or master students, but I had them build a panoramic scanner, a mechanical scanner that worked like this. It had a mirror that wobbled vertically while the assembly slowly rotated horizontally across the field of view. There was a question of mounting the camera on a tripod, and they were discussing what size screws to use. One of them said, “I think we should use 3/8 inch screws.” I said, “Do you know you can hold up an automobile with a 3/8 inch screw?” It was a big surprise to them. When I went there, Columbia was admittedly a very old fashioned place as compared with MIT for example. I took courses in strength of materials, thermodynamics and things like that. My first wife got an engineering degree from Cornell. She even took courses in how to deal with concrete. In India, where I was teaching for two years, they still teach all of that to undergraduates. I am convinced that one can have a modern engineering course and still have graduates with some knowledge of the tangible world of constructed things.

Teaching and research career highlights

Nebeker:

I wanted to ask you, I know we have already gone on for quite a while. You have talked about quite a number of things, but I’m sure they’re very important bits of work that you have done that we haven’t mentioned. If I could just get you to name a few other things that occur to you that you are proud of.

Schreiber:

I recall that as we went through that there were one or two things I skipped right over.

Nebeker:

Well, I mean one thing that I also wanted to ask was which of your students have maybe gone on to be most recognized?

Schreiber:

Well, Tom Huang is the one who has gone the furthest in academia, being a well known professor at the U. of Illinois. He was one of my first students. A few have become academics, not many. Many have gone into companies to make things and do very well. Of course, they wouldn’t have come to me unless they were so inclined in the first place. One of the interesting things that you may not want to put into your report is that on the television project, more than half the graduate students were Asian or Asian-Americans. I attribute that to the fact that these kids all know each other and socially they hang out together, and when one finds something good to do his friends come. So that’s part of the explanation for what happened. Another part is that Asian parents are involved with their children’s academic programs to a greater extent than is common in the US. They urge their children to pursue programs that they think will lead to higher prestige and earnings. I became aware of this during my stay in India.

When I was in India I got along very well with the people there, many of whom remain close friends. I identified with the culture. I learned a lot of Hindi, even the script. (The Indians appreciated that; they assured me that I must have been a Hindu in my last incarnation.) India has a lot of serious social problems, but I just liked the way so many things were approached. One is the parents’ great concern with education.

Nebeker:

For example, we haven’t talked at all about your television work. Maybe you just want to say a few words about that.

Schreiber:

About the HDTV business?

Nebeker:

Yes. For the last 15 years I know you have been heavily in…

Schreiber:

I talked about that this morning at the meeting at great length. I have learned something about how to manage R&D. One of the things I did to prevent the students from getting emotionally involved in the wrong approach (a very common failing that was partly responsible for the Challenger disaster) is that at the beginning I absolutely forbade them to invent television systems. I said the first thing we’re going to do is to study what other people are doing. We’re going to look at the problem very carefully, try to set up specifications for something that would be better, and then and only then would we start thinking about specific systems. I didn’t completely succeed in convincing all the students to do that. One of them was a key person in the development of ACTV, the NTSC-compatible system pushed by Sarnoff. I took him aside at one point and said, “Don’t get your career totally tied up with this system because its success or failure is out of your hands; you’re not going to be able to do anything.” Actually, by that time, I had become convinced that such systems were the wrong way to go and eventually the FCC came to agree with me, against the overwhelming opinion of the TV industry.

Mostly I have had very good relations with my students. There were a few with whom I didn’t. We didn’t get along for one reason or another—you know, they didn’t feel I paid enough attention to their ideas. I tried to let them have their head way as much as possible, and yet keep them on the track toward something that I thought would be an appropriate goal. I didn’t do that perfectly. I had a few failures, but my very last successful PhD student was finally a woman. (Women in EECS at MIT have had a poor environment, but it is improving.) She has a lot of self confidence, and is a good student who did a terrific thesis. She has a marvelous job at HP at an astronomical salary. It just wowed me how much they were willing to pay. She turned down an even bigger offer from a Japanese company. All Japanese companies established plants and labs in the US to take advantage of the cutbacks in R&D and the resulting surplus of highly educated researchers. For example, when GE bought RCA and spun off the Sarnoff Laboratory to the Stanford Research Institute, the first thing SRI did was a 25% job cut. A large proportion of the good people that were let go at that point went to work for Japanese companies. It was awful.

IEEE Signal Processing Society

Nebeker:

Another thing I wanted to ask you about was, of course I have been looking in the last year or so at the way signal processing including image processing has developed and how the professional society, which earlier was audio and electric acoustics, how it redefined itself. How do you think image processing has been served by this professional society? This Signal Processing Society?

Schreiber:

I don’t think I’m in a position to give a sound opinion on that. Of course we now have lots of meetings on image processing. This meeting is an example. Such meetings are certainly very good at spreading the word, and if you get the right speakers there is an important educational aspect. People find out what other people are doing and more important how they are thinking about the problems. That’s good. You have to have at least some of that. However, I have the feeling that we now have too many societies, too many meetings, and too many publications, but I don’t know how you can control that.

Nebeker:

For example, when the Transactions on imaging processing was established, I know that the computer society argued that such a publication wasn’t needed. That there was their Transactions on pattern analysis or whatever it was called. I’m just wondering if things that have happened have appeared to you unnecessary or missed steps?

Schreiber:

As I said, we have too many journals, too many meetings, too many papers being published. Almost anything can be published now if you send it to the appropriate journal. When there are so many publications they are all looking for material, and it becomes a sellers’ (authors’) market. If you learn what kind of thing the various journals like, it’s easy enough to write a paper that will be accepted, which certainly was not the case when I started out. Part of the reason we produce too many papers is that there is a tendency in the United States of researchers failing to look at the history of what they are doing. Europeans are quite different in this respect. They tend to read the literature much more carefully. In India, where I spent two years, people think they are really isolated, so they read the literature much more carefully than almost anybody else.

Nebeker:

You mean past publications?

Schreiber:

Yes. So they are aware of what other people do much more so than Americans, who tend to reinvent the wheel.

Nebeker:

Anything else you would like to comment on?

Schreiber:

No, I think that’s it.

Nebeker:

Thank you very much.